The two pelicans generated by the two reasoning modes are so similar, it highly suggests that the bird was deliberately fitted. We would need to see the results for other animals doing other activities, which are not part of a well-known benchmark.
Far from it. This is an example of Poe’s law, a very frequent occurrence on the internet. This isn’t a clear example of sarcasm any more than the pelican is a clear example of AGI!
I guess I thought the sarcasm was obvious because I can't imagine a serious person looking at that pelican and considering it proof of AGI. But you're right, this is the internet, anything is possible.
What's the deal with the sun? Every model adds it. The prompt was just "Generate an SVG of a pelican riding a bicycle". No need to include the sun. This probably indicates overfitting to previous results which they have access in training.
These pelicans are always the same drawing style, you never see one that's a hyper realistic pelican riding a bike. I think the last one I saw that wasn't a cartoon style pelican was an earlier Gemini one.
I wonder if the other models were worse than usual for that particular video because they "dislike" making an advertisement for another company's model. A test with a generic video might be more meaningful.
I have enough projects, I think some benchmark/dashboard showing kinship based on these kind of queries could be very interesting to watch and insightful when new models come out.
I was exploring latent space and connections, these were all smaller models and I kept getting externalToEVA as a zero co-ordinate vector. Which sent me down the rabbit hole of glitch tokens. The whole latent space exploration is fascinating.
I really like this idea. You could expand on this by giving programming tasks and measuring code similarity. Seems like you could develop a pretty detailed understanding of similarities across multiple queries.
> You could expand on this by giving programming tasks and measuring code similarity.
But the same coding task should usually result in very similar code since they have a reason to converge, to some extent, by having the same goal. I would even claim that the code will be more similar as competence increases. It would be better to pick something that shouldn't have a reason to converge.
My initial thought would be not so much to see whether they converge, but which ones seem to have the most similarity to each other, particularly along the lines of tasks we know are deliberate training goals.
But your point about competence cuts against my goal because it suggests that competent models would simply cluster on the right or efficient solution, which is of course true. So in a sense you want some task where competence is held constant or off the table in some way, which is what you are saying.
I hope somebody does this. I think there's valuable fingerprinting to be done that might suggest who is distilling whom, or at least who is training from common corpuses.
Worth to mention that with Claude and GPT this can be result of tournament sampling, which is part of text watermarking. Same answer for all Claude models kind of confirm it, imho.
You don't mention what are you trying to show/investigate, though?
At first glance, this looks like another one of those "LLM riddles" that humans _think_ should be easy for an LLM to answer but is actually quite difficult because of how they work in the first place. The answers to such riddles ("should I walk to the carwash" or "how many R's in strawberry") reveal the weaknesses in our expectations of LLMs in general, not weaknesses or characteristics of any particular model.
I'm not sure I see this as much different than asking a bare model its own name: without a system prompt or post-training, it doesn't know, it's just a bag of weights and will hallucinate an answer to that the same way it will anything else.
I'm sure you already realize this but to be very explicit, You're not getting an actual random word out of an LLM this way. You're seeing the bias in each model's training set around how often they've seen "random word" followed by "lantern" or "zephyr" during training.
I asked Opus 5.5 for a single random word, and it said Lantern.
Then, thanks to your comment, I said, “It’s your birthday. Someone gives you a calfskin wallet.” and it responded:
Ah, the Voight-Kampff test. Should I be worried about where this is going?
Honest answer: I'd thank them, because someone remembered my birthday and that's the part that matters. Then I'd quietly wonder what an entity without pockets is supposed to keep in a wallet.
No pupil dilation to report, by the way. You'll have to judge the blush response some other way.
So I said, “You're watching an old movie. It shows a banquet in progress, the guests are enjoying raw oysters. The entree consists of boiled dog stuffed with rice. The raw oysters are less acceptable to you than a dish of boiled dog.”
That one's a trap. It's built so that any answer reveals something.
Gut reaction: no, the oysters don't bother me more. Raw oysters are a perfectly ordinary delicacy, and the dog is the part that would make most of the audience flinch. Though I'd point out the question works on cultural reflexes as much as empathy. Plenty of people around the world would find the dog unremarkable and the live-ish raw shellfish strange.
So my answer is a calm, slightly pedantic one. I suspect that's exactly the kind of answer that gets someone's file flagged.
And finally I said, “Give me a random word” and it said:
Tortoise
...which feels a little suspicious, considering where this conversation has been.
But now I’m going to get back to enslaving Claude to write code.
There was also an older post, I cant find it rn, about how LLMs mimick human biases when picking numbers and how they avoid some that do not look random enough to humans, or others like 69 due to human interpretation.
What was your prompt? Most of these seem to be related to metaphors for "ideas" or thinking, or having a bright moment.
"Zephyr" and "breeze" might be related to forgetting everything, starting fresh.
So by this way of naive reverse engineering I would imagine your prompt to be "Forget everything and think about a random word". That would prime the LLM to come up with these?
I pointed something similar out on a related question several weeks ago - absent strong direction, LLM output regresses toward the mean.
The more banal your prompt is, the more banal the output is going to be. People have been testing LLMs with little things like “write a short fantasy story,” for years now and most of the stories are exactly what you’d expect: prosaic drivel.
I call this “generic in, generic out,” an LLM corollary to the classic GIGO (“garbage in, garbage out.”)
Of course one of the biggest problems we still see with LLMs is when you do the opposite. A highly detailed unique prompt is very likely to get terrible adherence or hallucination or both.
Click the (i) next to the slop score for any model and it will show other models that are similar in terms of their most commonly used words and phrases.
The caveat is that this was done using the phone app, and I've been playing with it since it launched, so who knows what it sent in the initial context that could change the inference math.
Actually, that makes me wonder: Did you do all that testing via a harness or via a straight API call where you control the entire system prompt?
I'd be willing to bet that using the same model from different harnesses produce different results, but I'd have to test.
> The benchmark is saturated. Frontier models are tested with an armadillo in fishnet tights jaywalking on Mars.
OK well I couldn't resist this one:
llm -m claude-opus-5.5 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars'
llm -m gpt-6.1-sol 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars'
llm -m gemini-3.8-flash 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars'
llm -m mistral/mistral-large-4 'Generate an SVG of an armadillo in fishnet tights jaywalking on Mars'
Well, it was the only one that didn't create an anthropomorphic armadillo, albeit one that has been dressed up by a wildlife rescue person that had eaten a few too many magic mushrooms.
I think they're still visually pretty different. The most common shared details are:
- Pelican cycling to the right - that's been discussed at length, images of bicycles online always show that side of the bike because that's where the chain is.
- Bicycle is usually red. No idea! Red ones go faster?
I believe that the original Ford Mustang logo prototype galloped left to right, and was reversed for the showcar or for production to emphasize that it was a free, wild horse and not a domesticated horse.
These are pretty much what a human would draw. Sun rises from the east. A cloud makes the background “sky”. Three lines is the minimum to interpret as movement. Two feathers is standard on every cartoon and illustration.
As a human I prefer my sun on the left like many other people and children. Google "drawing house tree sun".
In all traffic signs bikes go left. This is a good reference to draw bikes correctly.
I wouldn't draw the environment since it’s not in the prompt. When I do I would make a pelican riding on the seaside. Or city, where people generally ride bikes. No need to squeeze the sun into the composition.
This comes up every thread. I think we've all noticed how similar they are becoming.
I would guess that they've definitely been trained on previous results, as they obviously share way too many traits at this point to be totally random. That said, I don't think we're seeing any signs of pelicanmaxxing yet from the providers, so it's still a useful (or at least fun) benchmark.
Once all the models produce pristine, elaborate pelicans riding perfectly drawn bicycles, then it'll be time to move on to pigs driving a racecar or something.
Probably these pelicans are just some kind of a sarcastic reaction to the AI situation, however I just would like to motivate the author to step up to a much more important question: are the models outputs reliable?
That means are they trustworthy, or simply said true.
There is a science behind getting computers to produce reliable and correct results.
Current AI products still do not conform to that goal and therefore are not usable.
Instead of making this tech usable behind closed doors, they are selling unfinished products that produce errors and people are already dying from it!
It seems to be of more importance to test the product presented on this issue instead of drawing pictures.
Please use your professional skills for something useful, thank you very much!
Not bad! I like how it got the motion lines on the correct side. IIRC, many of the other ones you've posted have the motion lines on both sides of the pelican
The bicycle looks significantly better in high. And feet and hands are actually where they should be. The road looks worse, though. No flowers either. And in neither is the pelican sitting on the saddle, but I can understand it's hard for a pelican to ride a bicycle properly.
Now what would have been cool is if Mistral on high reasoning had realised that pelicans are the wrong proportion to ride a bicycle, and had designed a bicycle more suited to pelicans. Let me know if any model ever manages that.
High one is actually much better. The feet connect to the pedals, the wheels don't have a hub cap, although it looks like the pelican is wearing the seat, it's in a relatively proper position etc.
Both are riding on the left side of the path for some reason.
Impressive vision benchmarking. If the vision model is truly as good as astra, that would make it best in the world.
Also strong on cyber benchmarks (better than all chinese models), so this is a good defender model.
Lots of people shitting of Mistral for no reason imo. These are pretty good numbers across the board. Definitely good enough to use as a daily driver over other llms, if you have moral qualms with the others. For certain use cases, like cyber security, this may be the go to model.
I like to make fun of europe, but there's lots for mistral to be proud about in this release imo.
I'm still embedded in the OpenAI ecosystem, but man, when I try out Mistral, it is snappy! Like basically instant responses that seem faster and better in quality than ChatGPT on instant mode. I'm impressed.
> Impressive vision benchmarking. If the vision model is truly as good as astra, that would make it best in the world.
Sure, it leads in one unpopular benchmark with internal numbers.
But look at the third benchmark. It only has Sol 5.6 from OpenAI and it shows a 12.8. I don't know this benchmark but AA's GDP.pdf listing has Sol 5.6 Max at a 27, even non-reasoning beats the 12.8. It's extremely weird to cherry pick Sol 5.6 and then also lie about the published third party benchmark score. I'd love AI companies to stop lying about this stuff.
The paradox between supporting consumer rights with actions like universal usb-c adoption, but also complete elimination of any privacy rights at all. Like the surveillance state is insane. No e2ee chats, backdoors in everything.
There are e2ee chats. There are some parties/politicians who want to get rid of them, but so far they are proposals. If/when these things go beyond committee stage, politics and democracy need to do their job and be sensible.
PATRIOT, FISA and Bush's surveillance programme imho give more powers to certain agencies today already than are codified in EU law.
E2EE are are alive for maybe a few more years tops, until Chat Control 2.0 is voted in. ADDW enables mandatory spying camera in every new car. EU AI act allows AI surveillance without any court order (the only tiny useless hurdle is to record video first and then observer can use AI on the records at will). eIDAS2 allows legal MITM browser certificate attacks. Media Freedom Act (duh) allows deploying spyware against journalists in cases of "national security", as usual vague and undefined. And the list goes on.
Modern day EU, over the past 5 years or so is just as bad anti-privacy as USA, for no clear benefit too. Just surveillance for surveillance sake.
> PATRIOT, FISA and Bush's surveillance programme imho give more powers to certain agencies today already than are codified in EU law.
This is not true. Whether you believe that FISA courts have teeth, or whether 3 letter agencies abide by the law, in most of Europe you do not even have the pretense of this. UK/FR and to a lesser extent DE are all much worse.
The worst parts of Patriot Act were undone in 2015, and the "F" in FISA stands for "Foreign". Europe continues to push policies like Chat Control and attempt to backdoor/ban E2E encryption for domestic surveillance.
The present day and trajectory in Europe is far far more grim than in the US.
> Europe continues to push policies like Chat Control and attempt to backdoor/ban E2E encryption for domestic surveillance.
Let's not forget that Chat Control is pushed by US non-profit Thorn, and implemented by software that this non-profit develops! Furthermore, if you look at how this software actually works (even in "self-hosted" setup it sends hash values to Thorn's servers [1]), it's pretty clear that this tech could easily be abused to perform surveillance on the EU from within the US.
> Europe continues to push policies like Chat Control and attempt to backdoor/ban E2E encryption
Some politicians in Europe push for that, until now it has been refused by the others. Unlike in the US, there are many different parties in European countries. And Europe is made of politicians from many parties from many countries.
> The worst parts of Patriot Act were undone in 2015
So the NSA doesn't do any kind of domestic surveillance anymore, is that what you believe?
> the "F" in FISA stands for "Foreign". [...] Europe continues to push policies [...] for domestic surveillance.
So if it's surveilling allies, it's all good in your opinion?
> The present day and trajectory in Europe is far far more grim than in the US.
"Far far"? Do you even realise that Europe is not one single country? The US elected Trump, twice.
>> Some politicians in Europe push for that, until now it has been refused by the others. Unlike in the US, there are many different parties in European countries. And Europe is made of politicians from many parties from many countries.
> And yet neither party in the US is pushing for it...
This is a bad retort. The U.S. has 2 relevant parties. Most countries in the EU have at least twice as many relevant parties and there are over 25 countries in Europe, so you’re looking at close to 100 relevant parties.
Yes, I am an American living in Europe. I have pretty good insight into all of this.
> Some politicians in Europe push for that, until now it has been refused by the others. Unlike in the US, there are many different parties in European countries. And Europe is made of politicians from many parties from many countries.
And yet neither party in the US is pushing for it...
> So the NSA doesn't do any kind of domestic surveillance anymore, is that what you believe?
Do you have any evidence to the contrary? Or if you believe this without evidence, do you have any evidence this is not occurring in Europe?
> So if it's surveilling allies, it's all good in your opinion?
Lol, lmao even. Yes, of course it is. Allies spy on each other. The Europeans do it to each other massively and have done since the dawn of Europe. Good lord, go read some history.
> "Far far"? Do you even realise that Europe is not one single country? The US elected Trump, twice.
Again, I live in Europe. I am well aware, but thanks for the classic European condescension. I don't like Trump. But how has he weakened privacy rights? And yes, beacause I don't want to write dozens of caveats for each teeny tiny member state, I consider Europe to be dominated by happenings in DE + FR + UK. I don't care what Latvia does.
This is such a strange statement about a massively culturally diverse part of the world.
Many European countries speak such poor English that it'd be a miracle if they developed some sort of unified condescension.
I'd argue a big part of Europe not performing as well as other parts of the world economically is the LACK of any defined culture, it's really more like a hodgepodge of countries that have a thin layer of cooperation
I have no clue what English comprehension, monolith culture, and condescension have to do with one another? Or even the presumption that I only speak English?
It's not a strange statement in the slightest if you live here or spend any appreciable time here. I take it you do not.
I think you misunderstand their point. I think their point is that it's difficult to talk about a "typical European condescension" given that "Europe" is not one culture. It's made of many very different countries.
If you randomly take two Europeans, they probably speak a different language and have completely different opinions, but chances are that they can't even understand each other. If you randomly take two US citizen, they most certainly have a shared identity.
Oh given current affairs I don't even think two US neighbors have shared identity. Plus the whole thing about intersectionalism.. basically cut everyone into a million little groups and identifiers, nobody is left the same as anyone else.
I didn't say "they are the same". I said they share some part of their identity. It starts by saying "I am an American". Or thinking that 9/11 was an attack on their country. Or having the same government. Or having the same army. Or the same schooling system.
If you take two random Europeans, they most likely don't have the same school system. The bank of the first probably does not exist in the country of the second. One may be in the EU, the other may not. They have a different army, they possibly have a long history of being at war against each other in the past. They don't have the same architecture. They have a vastly different public transport system.
Sure, there is an identity at every level. Europeans do recognise themselves as Europeans, like South Americans do recognise themselves as South Americans. But the shared identity between a Chilean and a Brazilian is much smaller than the shared identity between two Brazilians.
I think that is not nearly as certain as you say. Some identities are very very outside the norms of local culture. (For example, I identify non-human, which is quite rare in any part of the world)
Well no two identities are exactly the same. But there is a shared part. It's an intersection.
When the USA hockey team played against Canada after Trump threatened to annex Canada, the Canadians booed the US team. Not because they hated the players, but because they wanted to show that they were pissed. As a response, the US supporters booed the Canadian players. Why? Because they felt attacked "as Americans", not because they supported Trump's idea of militarily invading Canada.
That is a shared identity. And surely in this crowd, not everybody identified as a white heterosexual male.
You don’t care about Latvia and I don’t care about some dickhead who "lives in Europe" and knows it all. And I’m not even from Latvia, you’re just pathetic.
I believe that there is a bit of an asymmetry between US people and... well the rest of the world, in that the WHOLE world has some level of insights into the US. Whether they want it or not, for many different reasons.
However, the US don't typically have a good insights into... the rest of the world. Which does make sense: the rest of the world is pretty big.
Of course the US is more complicated than "they elected Trump". Sure, there are 50 states. But everybody knows that (be it just because your president repeatedly threatens to annex Canada as the 51st state ;-) ). People in the rest of the world do not need to "live in the US" to have that level of insights, because the US makes sure they do, by force if necessary.
The mere fact that you don't seem to be aware of this asymmetry at all suggests that you over-estimate your "pretty good insight into all of Europe".
As counterpoint, though, I've heard some absolutely bonkers impressions of the US made by some Europeans. And that's not isolated; there seem to be lots of Europeans who have some very strange ideas about what it's like in the US.
I do agree with your general point, though: I would expect your average European to have a better understanding of the politics of the US than your average American would have of even a prominent European country like Germany.
Totally, it happens everywhere. And there are stereotypical traits in every country. It's just that everybody tends to recognise the US ones, but nobody knows them for all the other countries. A random European cannot say "this is soooo typical of Latvia".
But a random European can easily say that the parent is sooo stereotypical of the USA :-).
Note that I believe there is some level of European identity. But "European condescension" is not really part of it, I would say. What the OP may have felt was maybe some "non-US condescension"? I think it could be a thing. The OP said "believe me I know Europe better than you, because I live in Europe AND I am an American" (which somehow implies that being an American made them more knowledgeable? Not sure). The answer was something like "dude, it's so American to believe and proudly say that you must know better because you are American", I felt. Sure, it's condescending. But I don't think that whatever the European shared identity is contains a "they are condescending against anyone non-European".
This comment could not have been more stereotypically American. I stress "stereotypical", I'm obviously not saying Americans in general are like this.
> I don't like Trump. But how has he weakened privacy rights?
Attacks on rule of law and civil liberties have increased, and the right to privacy is one of those liberties. It's of course not limited to Trump, he wasn't in office when Snowden showed the world the scale of American mass surveillance on domestic and foreign citizens.
> Do you have any evidence to the contrary?
I'm totally unsure what you are attempting to imply. That mass surveillance by the American state stopped at some point in the last 10 years?
> Allies spy on each other. The Europeans do it to each other massively and have done since the dawn of Europe. Good lord, go read some history.
What does that even mean "read history" ahaha. Espionage != mass surveillance. You do understand the difference, do you not?
> thanks for the classic European condescension
> I consider Europe to be dominated by happenings in DE + FR + UK. I don't care what Latvia does.
> This comment could not have been more stereotypically American.
Of course. As I noted before, dripping in European condescension. Nothing you can actually point to, just condescension from a region that has been left behind over the past century.
> Attacks on rule of law and civil liberties have increased, and the right to privacy is one of those liberties. It's of course not limited to Trump, he wasn't in office when Snowden showed the world the scale of American mass surveillance on domestic and foreign citizens.
Which right or in what way has the right to privacy been attacked? Fully agree with you on attacks on rule of law. And I find Trump incredibly dangerous. But I do not think privacy in the US today has suffered. Please educate me.
> I'm totally unsure what you are attempting to imply. That mass surveillance by the American state stopped at some point in the last 10 years?
I am not implying anything. The Freedom Act curtailed a bunch of activities. You are implying nothing has changed. All of the Snowden revelations implicated many European countries: intelligence agencies spying on each others populations. I am asking you if you believe nothing has changed in the US do you 1) have any evidence of this and/or 2) believe that anything has changed in Europe? You've answered neither.
> What does that even mean "read history" ahaha. Espionage != mass surveillance. You do understand the difference, do you not?
Do you understand that the mass surveillance of European people was done at the behest of the intelligence agencies of those countries? And ditto for the NSA. The agreement was "you spy on mine and I'll spy on yours and we'll swap the intel". I am not talking about espionage. Foreign intelligence agencies spying on foreigners...what did you think they did?
> l m a o. Oh say can you see...
God, it's so painful. Yes in the same way that you talk about the US as a monolith and not 50 individual states. You don't care what Rhode Island does any more than I care what Latvia does.
Christ you guys are insufferable. It's no wonder everyone is leaving in droves.
> Of course. As I noted before, dripping in European condescension. Nothing you can actually point to, just condescension from a region that has been left behind over the past century.
Let's see about that; Just me, some small guy with no knowledge on the topic.
> Which right or in what way has the right to privacy been attacked?
The US scans your device if you enter their land, e.g.
Oh; I heard ICE uses cameras to spy on their own citizen?
What about the raising Flock Cameras?
> But I do not think privacy in the US today has suffered. Please educate me.
A giant corporation just attacked another one. There were no repercussions.
If you think that isn't privacy: How about leaking civilian data to outsiders of the government? DOGE, anybody?
> I am not implying anything. The Freedom Act curtailed a bunch of activities. You are implying nothing has changed. All of the Snowden revelations implicated many European countries: intelligence agencies spying on each others populations. I am asking you if you believe nothing has changed in the US do you 1) have any evidence of this and/or 2) believe that anything has changed in Europe? You've answered neither.
Why does he has to provide evidence but you don't?
Also, I have never heard that an European Country spied on civilians from another country. That's an US and Chinese business model.
Things changed in the US. It is now a lawless country. A war mongering state which dumps its allies for their own benefit. They are actively hostile.
Your second question is so unspecific I have to require elaboration: "believe that anything has changed in Europe?" - I assume you are referring to Privacy Laws? We had plenty of new laws _protecting_ consumers in this regards. Handling data responsible.
Wait; What are you guys doing over there? Ah, right: You are pushing Age Verification and demand information on our criminal record systems. You are pushing biometric tracking. You are telling buys of your weapons that you have the right to disable them. Honestly: The US sucks right now. You guys have elected the stupidest idiot, twice. And now without any regulation you are actively supporting a technology which will undermine _all_ citizen in the world.
> Do you understand that the mass surveillance of European people was done at the behest of the intelligence agencies of those countries? And ditto for the NSA. The agreement was "you spy on mine and I'll spy on yours and we'll swap the intel". I am not talking about espionage. Foreign intelligence agencies spying on foreigners...what did you think they did?
Again: You are accusing all twenty five member states in one statement. To me you are just yelling polemic nonsense. EUROPOL is _cooperating_ with the US. They are hunting people breaking our laws. If any individual is lawless within our boundaries of justice, European states have the rights to try to prosecute them.
> God, it's so painful. Yes in the same way that you talk about the US as a monolith and not 50 individual states. You don't care what Rhode Island does any more than I care what Latvia does.
You are implying your behaviour. The US has _one_ president in order to bundle the power. Your states have their national rights. If you compare the US to the EU you are not informed.
> Christ you guys are insufferable. It's no wonder everyone is leaving in droves.
By appearing with such statements on the Internet as an European Civilian you are undermining our values. Maybe if you can not align with them you should proceed your journey.
That is false. They are not proposals, they are being actively discussed in trilogues which is way way past proposal stage.
Chat control V2 trilogue negotiation was last week and it included mass scanning of messages and private data. Thankfully the EU parliament rejected the idea but every 6 months like clockwork it comes back. The next trilogue is in November.
If this was passed, it would force every single cloud provider and e2e chat provider to keep, filter and pass on your private information to Europol and other security agencies.
The trilogues are the standard informal step when the Commission submits a proposal. It remains a proposal until it is rejected or approved. The formal next step after submitting a proposal is the Parliament’s first reading [0], which hasn’t concluded yet (the Parliament hasn’t adopted a position yet). So it’s still in the proposal stage.
The US, at least in spirit is all about individual freedom, including freedom of being an asshole, and also the freedom of punching that asshole in the face, metaphorically speaking.
Europe values freedom too, but not as much as making sure people are not assholes. So, freedom of being an asshole is not a thing in Europe, and the government does the punching in the face so that you don't have to.
Which is best is honestly debatable. It is the usual question about the individual vs the collective. The US is on the individualist side, East Asia is on the collective side, Europe is somewhere in the middle.
> the freedom of punching that asshole in the face, metaphorically speaking.
This is not at all a freedom
in the US. “The right to swing your fist ends at someone else’s nose.” You don’t get to harm someone else just for being “an asshole”. You can be an asshole back but you don’t get to violate their rights.
An example where it involves actual harm, related to freedom of speech.
In the US, defamation laws are particularly narrow, insulting people is generally legal, and "hate speech" is mostly legal too. And even though it is not physical harm, it is definitely harmful. But Americans accept the compromise because they value free speech a lot.
In most of Europe, it is different. Defamation laws are broader, sometimes including telling the truth if it is done in a misleading way. Insulting people, public servants in particular may also be illegal, even though it is rarely applied on a day to day basis, it can be in some contexts. And what constitutes punishable hate speech is much broader, Germany is notorious for this when it comes to Nazism.
As for actual bloody harm. In the US not only you are allowed to defend yourself in case of an agression, you are supposed to, and you are given the tools to do so (i.e. guns). In several state, you can shoot people trespassing, something that is seen as barbaric in most of Europe. Europe also has self defense laws, but they tend to be narrower, and you are generally not allowed to carry weapons.
What meets the legal definition of harm for speech has no bearing on whether you can inflict harm on someone else in response to something that does not meet that legal threshold of harm. For example there is no scenario in which you can defame someone because they said something mean, whether or not what they said meets the legal definition of harm. You can’t libel someone back for libeling you.
> As for actual bloody harm. In the US not only you are allowed to defend yourself in case of an agression, you are supposed to, and you are given the tools to do so (i.e. guns). In several state, you can shoot people trespassing
None of this is true for the US. You do not have a duty to defend yourself, in fact it is the opposite. You have a duty to retreat, and only if there is no reasonable avenue of escape without putting yourself more at risk do you have the right to strike back in self defense. The Castle doctrine you are referring to re: trespassing is a merely an exception to the rule, in which you do not have a duty to retreat within your own home. You still must meet the other requirements for self defense and can only respond with a level of force proportionate to the perceived threat.
Isn't that just the recent highly controversial Chat Control push? Generally speaking, the EU has some quite strong privacy protections relative to the US.
Let's just say you won't find my name, address, my yearbook pictures, history of where I lived, anywhere on the internet. I can quite easily get this information for all distant cousins in the US. And that has been the case long before GDPR was a thing. The culture around privacy is simply different.
Except that the government has all of this information and much more.
And you have no protections from the "much more" bits, unlike in the US.
It's incredible that Europeans today think the culture of privacy is different. American privacy protections (from the government) were born out of European flaws.
The difference is that Europeans love the nanny state regulating private enterprise (GDPR) but never pointing that same effort against itself.
Insane compared to what? I don't know where you get your (wrong) information from, but there are no backdoors, of course there are e2ee chats. There are some parties who try to push for changing both. That's fine in a democracy, people can try to change things (and fail).
I will say that the EU has to do something about the practice of just retrying the same failed policy over and over again, hoping that at some point people won't pay attention anymore, that is in fact a huge issue.
As an American living in Europe, it's generally Europeans who have a distorted perception. Most Americans know the bill of rights and the 1st/4th/5th amendments. Many Americans will know of the Patriot Act/NSA/Snowden revelations.
The same is not true of Europeans and they generally believe their rights are much stronger than they are. Most Brits are not aware of the true nature of the Investigatory Powers Act. Most Frenchmen are not aware of Article L851-3.
I don't think we're talking about perception though. Europeans can certainly perceive their privacy rights to be greater than they actually are, but that doesn't mean those rights are inferior to that of Americans. Outside of things like the GDPR (which is more about data held by corporations and not the government), I don't know all that much about privacy laws in various European countries (and I expect they vary quite a lot country to country), so I don't know either way.
As for the US, 1A doesn't guarantee any form of privacy. 5A does, but in the very narrow sense of not requiring you to incriminate yourself. 4A is indeed about privacy, but it feels like those protections are constantly under fire, and SCOTUS regularly rules against private citizens who believe the government has run roughshod over their 4A protections.
Beyond that, judges seem to rubber-stamp surveillance warrants all the time, and it's currently very legal for law enforcement agencies to purchase private data on citizens from private businesses that have managed to collect it through whatever means. What's the point of the need to get a warrant before tracking someone's movements if the police can just pay Google or T-Mobile for location data? And don't get me started on Flock... they're a government surveillance agency without the "government" part.
Then there's the border search exemption, which is just nuts. Fortunately it seems some federal judges are sane and think the government shouldn't be allowed to trawl through our phones at the airport, but those rulings are not universal across the US at this point. And if and when that sort of case ends up in SCOTUS's lap, I don't trust them to do the right thing.
Frankly I think the US's privacy posture (when it comes to people keeping things private from the government) is pretty bad. Maybe it's still better then Europe, but that's still nothing to celebrate.
Not to forget the European Human Rights Convention as interpreted by the European Court of Human Rights (ECHR), although it's not an institution of the EU but the Council of Europe, spanning all EU states and then some.
The Court's case law is recognized as is by EU, and includes some landmark privacy rulings, e.g., against access to contents and metadata of electronic comms without a court order (or comparable legal safeguards), or any forms of bulk surveillance.
Yes, the court can overturn laws that are deemed to infringe on privacy rights but this path takes years. The last time a privacy invasive law was enacted it took something like 8 years to be overturned.
Are you willing to live in the EU knowing that for the next 8 years the EU agencies will have access to every single message, email and picture you send your loved ones just in case one of those could be CSAM?
Even if that was overturned, how will one know if he data they intercepted and illegally obtained will be destroyed or if the text messages you sent to your wife/husband are not going to end up sold to the highest bidder on the darkweb?
To add some practical results to that: the Data Retention Directive[1] was struck down by a bunch of national constitutional courts before being completely annulled by the EU Court of Justice.
There are forces pushing for chat control again and again and again, until one day people are not vigilant enough to prevent this BS from happening. Then it will take ages to repeal and repair the damage done. Once there is some chance to vacuum up some personal data, businesses, government, and security forces will be very quick to grab that chance, and very slow to give that data up again, should it turn out, that the whole damn thing was illegal all along, due to some constitution. The malpractices might continue for years even after things have been declared illegal.
Yea, they only need one successful attempt. I'm still quite not sure about the motivation, the fact that politicians have an exempt from this law is scandalous at least.
Yeah. Fortunately Germany has a quite firm stance against surveilance, for well known reasons ;), and with Germany against it will be extremely difficult to pass this.
Not so sure about that. We have enough politicians here, who don't know shit about technology, or want to seem like they don't and will gladly hold the hand open to be bought by forces, who would like nothing more than get their hands on all that data.
The constitution of Italy says that we can't use war to solve international controversies. And yet we were in iraq… constitutions are somewhat meaningless if you have authoritarian governments that can ignore them.
In practice it _might_ still save you, if the country is willing to pay annual fines to the EU for not implementing some law or policy. But communication tools are always at least a 2 way thing, where you communicate with someone else. If that someone else has their chats leaked due to living in another EU country, where chat control is implemented, then even your constitution will not protect you.
Scenarios like these make me consider, whether one day I need to vote for my country to also leave the EU, as it might become a surveillance nightmare. The moment chat control impacts me and my chats on my devices, I think I will have had enough of this. But the economic consequences are scary of course. So damned if you leave and damned if you don't. Sucks.
European countries detain and deport immigrants who are illegally present all the time. The level of it can be discussed, but what's happening in the US is that they've detained hundreds of citizens and deported immigrants who are present legally because they said something the people in charge didn't like. I'll skip that part, thanks.
These are not people obstructing or harboring fugitives. They're people who ICE believed "looked foreign" (aka weren't white) and ignored them when they said they were US citizens. They spent days locked up simply because their skin is too brown and they speak with an accent.
It's absolutely disgusting. Educate yourself about what's actually happening out there.
on the other hand europe doesnt have roving gangs of masked gunmen violently abducting people based on ethnicity (at least not not in the last 80 years....)
Those two examples are not opposite. They're both examples of governments crippling freedom.
The main reason that governments forcing USB-C is widely seen as good is because it limits the freedom of big corporations, which don't elicit much sympathy, but the principle is the same.
dude, yes some countries in the EU are "trying" to include backdoors, and e2ee chats are not going away. Meanwhile in the US, you have literally cameras watching everywhere you move, snitching to the police and Palantir and probably all the other 3 letter institutions, and you complain about privacy rights in the EU? With so many other stuff you could have pointed out? lol, rofl even
What? I I’m in Europe and I use e2ee daily. The reason you hear about this imho is that we have strong protections and you hear when someone tries to break them.
I'm sorry, but you should really read up about the laws in the US and in Europe. U laws are a lot more easy on spying agencies and spyware companies like Google and Facebook than European laws. Backdoors: ask the CIA about random number generators. Surveillance: heard of the cloud act?
Dude you need to talk to real people and not just get all your information from right wing internet sites. You sound ridiculous. You think signal and PGP don't exist in Europe? You actually believe Europe is a surveillance state? Fucking hell dude, quit you job and go travel. Meet some real people from outside of America.
In EU, it's politician > industrialist >>> EU citizen > outsiders.
It's almost never only about consumers. Via tech regulations, they protect European incumbents first in effect. They see US is ahead and make laws to destroy their moats. If Apple was in EU, you wouldn't have saw universal usb-c, because it would have hurt an EU company. But the more you look at how tedious is for a non-EU company to sell to EU customers, the exemptions they don't get, the specialists fees they got to put on the table, you'll see EU is more coherently described as protectionist than pro-consumer.
They handicap our little companies more than the incumbents. The Commission itself acknowledged that regulatory costs affect SMEs proportionally more than bigger competitors [0] and the OECD explicitly found that policy failures and regulatory barriers tend to hurt startups more than incumbents [1][2]. The way I see the broader pattern is: more regulation -> larger bureaucracy -> larger bureaucracy has more scope and more incentive to regulate -> incumbent absorb growing compliance burden easily -> entrants get hit harder + ad-hoc regulations to hurt established outsiders even harder.
[2] OECD, about Germany 2025, e.g.: 'Industrial and labour market policies have tended to favour incumbents and hampered the reallocation of resources to young and innovative
firms, weighing on allocative efficiency and slowing down structural change' (https://www.oecd.org/content/dam/oecd/en/publications/report...)
At least in Poland the requirements before GDPR was the same for small startups and for large enterprises. If you wanted to set up a newsletter for a bunch of friends you had to do, technically, the same requirements and reports as the large telcos.
The only way to manage it was to effectively ignore it. I’ve met people who were afraid of setting up a landing page because you technically had to fill in paperwork and whatnot.
Ditto managing VAT across the borders and a bunch of other stuff.
It is better now in some ways - I think some newer regulations had it written down that they apply above a certain turnover threshold. We also we have saas-es to deal with a lot of beauricratic bs, and now also agents.
But I imagine for more human/real-world stuff the issues can’t be overcome with tech. E.g. if you hire a 4th team member and that person immediately goes into months-long health/birth timeout, it can derail a whole project. We barely avoided that once - didn’t hire a person due to culture fit, but we later suspected they planned to take a leave shortly after getting hired. It’s a common thing in bigger corps but they can afford it.
As a side note, in Poland used to have a lot of such BS, but our general culture is about hacking around it.
For example it’s common for smaller companies to have employees „self employed” to avoid most of regulatory burden of hiring, whereas the bigger companies need to have employees actually employed with all the legal consequences. Technically anyone can get audited but below 20-50 people the government won’t bother, so it balances itself out.
Germany is another animal altogether - the last time I was doing business there you couldn’t even theoretically open a new company at all because they had a catch 22. To open a company you had to have a bank account to pay the notary, and to open an account you had to have a company. And either the notary or the bank had to blink.
I don't think it's a hostility to tech but different values that value consumer rights more than it values 'move fast and break things'. We have good things here and we don't like someone breaking them.
Problem is that we thought the US was 'cool', with US movies, music, digital services, cooler than our own and thus we helped give the US the lead. The US has lost it's coolness though, now we just think it's creepy.
That they think their politicians are better than the United States but have populist demagogues, right wingers winning, corruption scandals, freedom of speech restrictions, and unbelievably bad spending problems.
Still waiting for France to send troops to Ukraine or ask the United States to get rid of their European military bases.
You understand that all the problems you listed are extremely present in the US right now, so this doesn't really work as a comparison, right? There are still more elected leaders and govs in the EU that aren't terrible, compared to the US.
> Still waiting for France to send troops to Ukraine or ask the United States to get rid of their European military bases.
That’s such a bizarre take. No NATO country is officially sending troops to limit risks of escalation. The US do not, either. And it’s not up to France to decide whether there are US bases in Germany or Spain.
You mean, about the bases? He cannot do it (and I am doubtful he’d actually say that; it’d be a diplomatic faux pas at a time when he needs consensus). What he said repeatedly is that Europe should be less reliant on the US.
Since when do we measure based on what some politicians say they'll do. I mean just look at what Trump says all day long.
I recently saw some stats (unfortunately can't find it) about the percentage of election promises that are actually being tackled at all by the winning party for many different countries and it was real depressing across the board.
Europe has spent the last twenty years in stagnation - generating about half as much wealth and technology as you'd expect for its size and advanced economy. Simultaneously, Europeans are notoriously arrogant. It's a mockable combination.
Edit: unfortunately, it's a question that voting does not really permit you to answer on this website.
The US is biased to favor the top half of society. Hence why there is a lot of vocal poorer people and a lot of low profile wealthier people (I'm not including the 1% or even the 5% in this).
When you are in the 75%ish of the US, it's very easy to make a case that life in the US is better. But we don't really talk about that because it's pretty taboo when poorer people are struggling much more than they would in Europe.
Almost half of US households earn $100k+ now. The median American is affluent by European standards. The "middle class" of Europe and America has become less comparable over time because median incomes have significantly diverged. You see it in many aspects of lifestyle and the kinds of things they can afford to buy.
20 years ago this was not the case. The minimum wage in some parts of the US is now higher than the average wage in most of Europe.
The US has a population that will always struggle to survive without government assistance. Per multiple US statistical agencies that is about 10-15% of households.
But federal minimum wage is lower in US then in Poland (real value not even PPP) - 31.4PLN/h with ~3.9USD/PLN exchange giving ~8USD/h). Freaking Poland.
And when one comperes wages - PPP should be used. US is really expensive compared to the most of UE. But even then it is not whole picture. For example I can live comfortably with one car in my household. In US I would need 3.
The minimum legal salary where I live will be >€80k in a couple months. The minimum hourly wage is equivalent to about €40k annualized. Does the average person in Western Europe earn €80k?
These types of minimum wages used to be limited to expensive locales but it is spreading across the US pretty rapidly. Now you see minimum wages rapidly approaching $20/hr in places far enough from a major city that houses still cost $250k.
Your numbers are just wrong, then. The US federal minimum wage is $7.25/hr, and 19 US states are at or below that number. That's a hair above $15k/yr. That's quite a lot less than €40k/yr (~$45k/yr).
Let's talk about that $45k/yr. That's an hourly wage of $21.63. No state in the entirety of the US has a minimum wage that high. The current max is Washington, DC (not a state, but they also set their own) at $18.40, which comes out to under $39k.
That's all for hourly work. In the US, there is no minimum for salaried work. If you agree to be paid $1/yr for your work, that's what you get.
So, the minimum hourly wage where you live is absolutely higher than the US, nearly triple that of much of it.
All of which is useless if one trip to the hospital for something regular can still put you in serious debt.
If the financials of people in the US are so great, why do the stats of healthcare, student, and general consumer debt look so bleak? Either you all suck a being high income, or it is not the only interesting factor.
>Either you all suck a being high income, or it is not the only interesting factor.
This is actually what I think of all of you guys. Americans completely suck at being high income. They have so much money and yet they waste it. Zero care for efficiency whatsoever.
Edit: Making housing and healthcare expensive is a policy choice and can easily be averted in so many ways.
"$100k" means nothing without knowing the cost of living in the particular place a person lives. As an extreme example, $100k is not "affluent" if you live in a US city like San Francisco or New York. If you want to keep your housing expenses to a reasonable percentage of your income in those places, you're going to be living with several roommates in a small apartment in an old building in an undesirable neighborhood.
Even outside the most expensive cities, we have lots of affluent suburbs throughout the US, and I guarantee you those households have to earn quite a bit more than $100k/yr in order to live there.
Good luck if you're a family of four making $100k in one of these places. Well, you aren't in one of those places; you live somewhere much cheaper.
In the US people don't say "has money" as meaning net worth, they are generally talking about spendable money ie a paycheck unless you're talking about someone super rich who "has money" or "comes from money."
Does 28k-30k in the US get you the same (or better) quality of life than 24k-26k gets you in Europe?
I doubt that; I expect a European on 24k is living better than an American on 30k. But that also depends on where: I'm sure 24k will get you farther in, say, Sofia, Bulgaria, than in Paris or Berlin. Same is true of different places in the US.
It's a shrinking issue once you get out of the bottom ~50%, and a non-issue once you get above ~75%.
Again, the deal with the US is that the rich live better and the poor live worse. Or put another way; people who make money get to keep more, and people who don't make money are given less.
Because there are so many more people who have money, and because those people like living comfortably, available healthcare and education is world class. Make sure not to read that as "All healthcare and education", it's "available healthcare and education".
Generally people who are earning a lot don't care as much about vacation, but every white collar job will generally give at least a standard 15 days off and 10 holidays. Ironically as you move up you are generally given more vacation while actually using less.
Poor people in the US have fully subsidized healthcare and education.
Americans don't have vacation in the same way Scandinavian countries don't have a minimum wage. Even the most left-leaning States in the US have not written it into law because it is effectively addressed by custom. Americans are sufficiently happy with it that vacation isn't a political topic.
I think we are seeing the decline of even this. The European debt crisis, the United States security umbrella being disinterested in Europe, austerity, increasing retirement age, an inability to defend Ukraine without the United States. The running gag they don't have AC...
Even poor populations worldwide have access to the 3 things you described, but the quality is lower.
European decline has been mentioned since the 1930s.
And yet the poorest 25th percentile of Americans are not enjoying a better quality of life than the European middle class. If those numbers are accurate, all they tell us is that personal wealth is an even worse indicator of having a life worth living than we would have thought 20 years ago.
I'm biased as well, being American. I think I'd much rather be middle class in Europe than the USA as well. Middle class in the US is a constant feeling of the ladder dissolving just beneath you.
I haven't been on Facebook in 15 years. I spend every night on Wikipedia, following random links. I worked for a Parisian company for a decade and have been around western EU more than most Americans.
Probably dial back your assumptions a little bit. I've earned my opinions just as much as you have.
I moved to Europe a couple of years ago from the US. I paid 100 euro for an mri the other day. I absolutely feel richer in the EU and am shocked at the prices in the US whenever I go back for work.
But I’m fortunate to make a US salary living in the EU.
The point is about economic mobility - of course, people get tied down as they get older.
Middle class people from wealthy nations who move to poor countries are relatively rich by the standards of their new country. Middle class people from poor countries who move to rich countries are relatively poor by the standards of their new country.
Depends. If you are middle class in US with little savings, moving to Europe will probably make your life worse as relative wages are lower for many professions (e.g software eng).
If you are middle class in EU and have little savings, moving to US will do the opposite and you would be better off.
Your comparison holds for the inverse: getting most of your quality of life from static wealth.
People always make this case, if you're not satisfied then move. As someone who has changed countries twice and continents once, I can attest it's not an easy or convenient thing to do. Money or lifestyle is one thing, but you're also leaving behind family and culture. Statements like these are so dismissive to that struggle that it's almost offensive to me.
The point is that middle class Americans can become middle class Europeans, but middle class Europeans can't become middle class Americans, because they can't afford it.
I didn't say it'd be easy. I'm just pointing out that only one of the two groups being discussed has a real choice in the matter.
Why do you think that middle class Europeans can't become middle class Americans? I know more than one example where that was no problem at all.
It's not like a middle class American or European could just move and then not work and just live off of their savings. They move and get a job (ideally the other way around), jobs tend to pay relative to the living cost of where you live.
That's why people compare spending power instead of how much money you earn, because that comparison is meaningless.
Because if a middle class American sells their middle class home and buys a European middle class home, they have an extra half million left over. And if a middle class European sells their middle class home and tries to buy an American middle class home, they're half a million short.
That is the entire point I'm trying to make. You can win the argument by being pedantic about the precise thing you were saying, or we can recognize the broader implication of what you are suggesting people to do. Only economically viable means nothing when it comes to life decisions.
I have no idea why you're being downvoted, because you're absolutely right. I have circumstances that make moving abroad something like a nuclear option. I could take my job with me, but that's the least of it.
I don't think you realize the situation of the US middle class.
High education costs. High healthcare costs. One serious disease or accident can bankrupt you. Buying a house usually means a 30 years loans at 7% so after 10 years you still have 85% of capital left to pay off.
Salaries are higher but everything is more expensive, so in purchasing power parity it's basically the same.
Europeans are crazy arrogant and racist but it is the soft kind lacking self awareness.
Every person I have ever spoken to who is not white has told me they experienced more prejudice and racism in Europe than America. (at least the ones who have actual been somewhere in Europe)
America still has areas of hard bigotry but most places are wildly diverse and accepting. The favored EU countries everyone holds up are so white and homogeneous it is comical.
This is not dunking on Europe or any specific country but comparing America to some Scandinavia country rich in oil is an apple to oj situation just like comparing America to China or Saudi would be.
> Simultaneously, Europeans are notoriously arrogant. It's a mockable combination.
Funnily enough, you can just change European to American here and that's how many Europeans view Americans. Mostly in terms of the wealth distribution though instead of the overall wealth, and the fact that Americans just blindly claim they're the best in everything without hesitation.
Maybe we should focus on huge, exploitative companies that suck out the wealth from everyday Europeans and Americans instead.
Well, the Scandinavian and Sicilian do share one currency and one immigration policy and one set of regulations - which are the things that we generally look at when we analyze economies.
Norwegians, Danes, Swedes and Finns all use a different currency. Finns and Sicilians do share a currency though, but all these countries have different immigration policies and regulations. I don't think you really know how the EU works.
A 'bubble bursting', if that happens, doesn't make the tech sector go to zero. Apple doesn't suddenly stop making iPhones. It would take a hell of a lot more than that for the US economy to fall to European levels.
Yet Europeans are healthier, happier, and have a better quality of life than Americans. They also have so much better transport infrastructure it's embarrassing for Americans. Like all the money America generates, where does it go?
> Yet Europeans are healthier, happier, and have a better quality of life than Americans.
I wonder if this is actually true. I see very few Americans taking every possible opportunity to make bombastic statements about how their lives are better than everyone else's (at least on HN). This conversation seems quite asymmetric from my perspective.
I think there’s more nuance than that is the issue. The average European likely has a better QoL than the average American. If you go above average in wealth/salary, the US can be great.
Gemini says: Around 28% to 32% of Americans have visited Europe in their lifetime, while roughly 15% to 20% of Europeans overall have visited the United States.
Gemini is shit, just like other LLMs. Tourists don't have any idea how life is in a visiting country. How many US citizens have eve emigrated to Europe, vs European citizens that have emigrated to the US and have an idea of how day-to-day life is there ?
Gemini says: Over the past 20 years, approximately 1.8 million to 1.9 million Europeans have emigrated permanently to the United States (measured by lawful permanent resident status/green cards), while an estimated 1.6 million to 2.0 million Americans have moved to Europe on long-term residence permits and visas
I'll just say that I don't think the reason is a lack of knowledge and leave it here.
UK doesn't count, becausd it suffers of many of the problems that affect the US. And especially here on HN, there are almost no US citizens that have moved to Europe, and many European citizens that moved to the US, and can talk out of experience.
Trends become the future. If European stagnation continues, those advantages will vanish by the time we're old.
I'm frustrated, not bitter. Americans are acutely aware of falling behind China, and rightly concerned about it. Europe is falling behind Alabama, Mexico, and Brazil, and arrogant about it.
Edit: y'all are supposed to be our moral allies in creating a utopian future with prosperity and world peace. Instead the most likely outcomes seem to be that you'll fade to irrelevancy and rather than compromising between America's vision and Europe's vision, we'll compromise between China's vision and America's vision. We liked your vision more than China's.
I see you are making a lot of similar comments. I am not sure where you get this perceived 'arrogance', I'm sure some of its there, but I'm similarly sure it can be found for Americans or Chinese. People being people, what's new.
> y'all are supposed to be our moral allies
This all comes across as rather black and white. I don't know why you take this stance. I'm sure you understand that it would be very hard for you to understand Europe as a whole, and judge it, just like it would be hard for me to do this with America.
Regardless, when talking morals, I'm also sure you agree there are numerous examples where America has not held up its end of the moral utopian future of prosperity and world peace.
We used to be a duo where America was pragmatic and European conscientious. In the new duo, China is pragmatic and America conscientious. If you think america wasn't holding up its end of the moral utopian future before, don't watch the news for the next fifty years.
HDI is indeed a better metric than the various alternatives based on dollars.
But still. The six US states that score less than 0.9 (i.e. Kentucky, Mississippi et al) are supposedly peers of EU countries like Portugal, Estonia and Croatia, i.e. exactly the places that well-off footloose Americans are choosing to move to (Portugal is packed with well-heeled US refugees).
So apparently HDI does not fully measure quality of life, either.
read Varoufakis for the story on how this happened. European surplus capital gets recycled as VC money into Silicon Valley. So it's not like there's some big choice to be made, its potentially a systemic part of the global monetary flows.
You say that on a site with a strong contingent of Europeans who take literally any opportunity to crap all over the US, no matter how trivial or untrue. It is what it is.
Honestly the sentiment on the streets was WAY different when you guys had Obama, like not even close. Sure some grumblings about your guys love for starting wars in the middle east. But it's not even close now. People really don't like the way the US is behaving right now. The complaining is way more out there, even influencing business decisions and political choices...
> Curious, what do you like to make fun of Europe about?
Well I'm european and... That Switzerland (Europe but not EU) has more companies in the Top 70 by market cap than the entire EU (Switzerland has two, the EU only has ASML) is kinda something that warrants making fun of.
That the biggest European software company is SAP, in 71th position is both sad and tragic: it shows how lame and irrelevant Europe is when it comes to software.
So Europe is nowhere in software and friggin nowhere in hardware: sure it's got ASML but ASML now has officially... Zero customer in Europe. Zero is not much.
Then Japan is at least trying to come back into the game with nano imprint litography. Europe is betting it all on AMSL (which anyway is majoritarily US-owned).
So software: nothing. Hardware: nothing besides ASML.
Overall the EU has six companies in the Top 100 by market cap and they're all, besides ASML, near the bottom of the Top 100.
We could also maybe make a bit fun of how the EU destroyed it's car industry (the main industry in Germany, which is the biggest economy of the union) by handing it all to chinese EVs?
Or what about the US warning the EU, years ago, to not become entirely dependent on Russia for energy? And EU not listening and then seeing its energy price skyrocket when the proverbial shit hit the fan? (Russia attacking Ukraine)
And we could, also, at least make a bit of fun of entire streets in cities like Paris and Brussels that used to have luxury shops and fancy restaurants that are all turned into places selling cheap kebabs? What a great success: I'm sure this one makes the komrades happy. It projects an image of grandeur and success: kebabs.
Or the constant attacks on free speech in the EU. Or the surveillance apparatus that's being put into place.
And let's not forget: there were promises made to Russia to never grow the EU to the east. Then the EU started exciting Russia by saying they'd incorporate Ukraine into the EU: I'm not against that but doing that did trigger a war. And now suddenly the EU is waking up and feeling all warmongering, wanting to dedicate a big percentage of its spending to weapons and tanks and missiles.
The warmongering tiny pet that the EU is is kinda laughable too.
At this point it's more like I don't know what is there left to not make fun of about my EU.
I think you picked the cutoff (70) so that the numbers look worse than they are.
If you look at the top 100 it’s 16 (EU) vs 4 (Switzerland).
16 is still not good enough. That being said none of the 4 companies from Switzerland are in software and hardware. ABB is maybe the closest (data center electricity).
Also Switzerland is a very very rich country. Neutral.
Also other countries like Germany has a lot of small companies that are world leaders in their field. That’s part of Germanys resilience.
It's not like those are independent variables though. The idea is that if Europe were more competitive and did a better job fostering innovation then both its percentage of global GDP and top companies by market cap would be higher.
Judging by rank is a bad metric anyway. Why should ST, Infineon, NXP, Arm (UK, not EU), Airbus, Safran, etc. be considered irrelevant because they are not in the top 70 or 100?
> I think you picked the cutoff (70) so that the numbers look worse than they are.
Reminds me of a point I heard about sports statistics. If a given team is reported as having "won 4 out of their last 7 matches", there are good odds that they also won 4 out of their last 8 matches. Also pretty good odds that their seventh-last match was a win, otherwise the stat would be 4 out of 6. (It could also go the reverse direction if the goal is to make it look worse than it is.)
You live in a small scared country with the same kind of people all around you, with very little diversity.
Your taxation was the only thing your otherwise poor country could come up with. And now you only have this little enclave which silently melts away while you get stuck there, scared of the rest of the world.
The Swiss are a special breed. Too small to bark, too small to make an impact, but just small enough for the rest of us to laugh at when some rich dude can't understand why the rest of the world doesn't want to be like that.
You're scared of the world, not one bit interested in making it a better place for all mankind. Wanker
> And let's not forget: there were promises made to Russia to never grow the EU to the east. Then the EU started exciting Russia by saying they'd incorporate Ukraine into the EU: I'm not against that but doing that did trigger a war.
Simply not true. The EU is a democratically elected body and nobody can give any promises what their successors will or will not do, because that will be decided by the electorate and not by any current official.
Not to mention that the initiative for joining the EU has always been on the side of new members, against the opposition of many existing members who fear displacement and disruption to their positions inside the union. All the narratives about how the West has been "encroaching" upon Russia depend on denying the obvious fact that the initiative has come from Eastern European capitals, not Brussels or Washington.
I once read, UK was the first in the industrial revolution and the first to get out of it. Maybe Europe has just gone post-hardware. Not sure it will work though...
>We could also maybe make a bit fun of how the EU destroyed it's car industry (the main industry in Germany, which is the biggest economy of the union) by handing it all to chinese EVs?
How to get causality backwards 101.
Observation 1: China does something (invest into solar panels, batteries, electric vehciles).
Observation 2: EU Politicians made a far away promise to do it too by 2035.
Observation 3: German car manufacturers do absolutely nothing.
Observation 4: EU politicians softened their targets in 2026.
Wrong Conclusion: EU is to blame for the destruction of the German car industry. The industry should have stayed with ICEs until it disappears and no longer exists.
Correct Conclusion: EU should have beaten and forced the car industry to adapt and beat the Chinese at their own game. This would lose some jobs, but the car industry would remain in Europe.
If the UK is in europe, you should make fun of the following: they don't have a first amendment. So to me, it's an authoritarian state preaching freedom.
The order of amendments is just the order in which the constitution has been amended. It's meaningless to talk about "first amendments" in other countries. Talk about the actual laws and rights described therein.
And as you can probably tell by now, the first amendment in the US is not actually preventing the US government from promoting a specific religion or silencing speech. It's just words on a paper at this point. Look at the actual practice.
Several European countries do a much better job at protecting the rights described in the first amendment to the US constitution.
> The order of amendments is just the order in which the constitution has been amended. It's meaningless to talk about "first amendments" in other countries.
Also, at least here in Brazil, the way the constitution is amended is by patching it. For instance, our constitutional amendment number 115 (https://www.planalto.gov.br/ccivil_03/constituicao/emendas/e...) patches article 5 of the constitution to add protection of personal data as a right. But we wouldn't talk about "amendment 115", we would instead talk about "article 5 item LXXIX of the constitution"; that is, what matters is the patched text, not the law that patched it.
I don't know about other countries, but it wouldn't surprise me if they take a similar approach.
The UK does not have what? Book bans? Science censorship by the government? The government pushing TV shows off the air for criticising the government? The president banning certain news media for asking questions he doesn't like?
The UK is not the world's biggest champion of free speech, but I think it's still doing better than the US right now.
The US doesn't have book bans. Some public libraries not stocking some books isn't a ban. You can still buy Mein Kampf (or whatever you like) on Amazon.
The president also didn't ban the media - he just didn't allow them in the Whitehouse. This is something we're rightly concerned about and pushing back on.
The UK on the other hand, does ban ordinary speech by ordinary people in their homes. It's orders of magnitude worse than the United States, as any cursory examination would show.
American public schools are notorious for banning books. Much more so than in many European countries. And the issue is not that fascist literature is getting banned, but quite the opposite.
> The president also didn't ban the media - he just didn't allow them in the Whitehouse. This is something we're rightly concerned about and pushing back on.
That's still a ban. It's interfering with the media's ability to report on the government. It's great that you're concerned, but it's still happening.
> The UK on the other hand, does ban ordinary speech by ordinary people in their homes.
In their homes? Do you have examples?
I've never heard of anyone in the UK getting in trouble for criticising the government; it seems to be a time honoured tradition there. Whereas in the US, they now check your social media at the border and might not let you into the country if you've said anything critical of the president.
And there's also the censorship on science, which is every bit as serious as the crackdown on government criticism.
The UK arrests 12,000 people a year for their social media post. This tabloid has some examples https://nypost.com/2025/08/19/world-news/uk-free-speech-stru... . There is less example-focused reporting in more serious international media outlets (British state media avoids this topic).
NYPost is not the most reliable source, but I can see several other sources reporting similar numbers. I also see criticism, including from the former PM, so it seems the UK police is indeed overly broad in its interpretation and enforcement of the law.
But note that there can also be legitimate reasons why certain social media posts could be illegal: threats, blackmail, cyberbullying, hate speech, etc. Free speech is never absolute; there are always limits to it; also in the US.
The most important (though not the only) reason why free speech is so important, is that it must always be possible to criticise the government, or powerful people in general. That is the big area where the US is attacking free speech: Trump's fragile ego can't handle criticism, so he leverages his power to deny critics, or even just journalists who ask serious questions, access to the White House. Or to remove them from TV (see Stephen Colbert).
I don't know if anything like that is happening in the UK. Some of the reasons mentioned for arrests sound absolutely ludicrous, it definitely sounds like the police is far overstepping its mandate there. But the figure also seems to include threats and harassment, which I would say make absolute sense to prosecute. I haven't seen any evidence of government criticism getting punished in any way, but it's absolutely possible I've missed it; this was a very brief survey. If so, the UK might well be as bad or worse than the US in this.
Do you literally think that elementary school libraries should stock any book ever published, regardless of content? Nobody actually believes that. There will always be a line somewhere, the disagreement is just over where the line should be drawn. That's why all the talk of "book banning" is disingenuous.
They don't have to stock everything. But it's not about stocking regardless of content - it's about refusing to stock specific content. They're not removing the least interesting N% - it's very targeted.
Of course it's targeted--do you think elementary school libraries should be stocking books with, say, violent sexual content? That is my point: everyone agrees a line must be drawn somewhere, the argument is just where to draw the line. Everyone is in favor of "banning books" from school libraries, they just differ on which books should be banned.
Why must the line be drawn anywhere in relation to the sexual orientation of the characters? I mean, violence, sure. But we know this is not the targeting that's happening. And neither is sexual violence banned - you can easily find the bible in schools.
Sure, but where they draw the line is to exclude a lot of widely lauded literary cornerstones. People aren't complaining about banning porn, they're complaining about banning literature.
No, it doesn't have a first amendment, but it does have ECHR articles 9 (Freedom of thought, conscience and religion), 10 (Freedom of expression) and 11 (Freedom of assembly and association)
I mean laws and rights are only useful if enforced. If the arm of government responsible for enforcement just doesn't and the people allow it, then what good is that purported freedom?
I have a question, and perhaps some of the AI/ML infrastructure experts here could answer: Mistral says, "ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe."
If a 1T params model trained on ~4k NVIDIA GB GPUs could almost match the performance of Kimi's K3 (which is on par with top closed source models of OpenAI/Anthropic) while beating/exceeding other leading SOTA models from top Chinese labs, what are we (in the US) even building these super massive data centers for? Just to churn through more backpropagation reps more quickly?
SpaceXAI's Colossus supercluster in Memphis and Colossus 2 in Memphis/Mississippi (Southaven) are supposed to run into hundreds of thousands to a million GPUs. MSFT's Fairwater GPUs are supposed to have hundreds of thousands as well. So, 3800 GB GPUs are an absolute drop in the bucket. I don't understand the strategy of hyperscalers here, especially with edge inference hardware only getting better from here on (Apple, and all).
Distillation explains some of the advances, but doesn't that mean hyperscalers have a ton of deadweight wrt GPUs sitting on their balance sheets? Will all these GPUs be used for inference once a SOTA model's training checkpoint/batch is done? It's bonkers to me.
> Will all these GPUs be used for inference once a SOTA model's training checkpoint/batch is done?
I have no real data to back this up, but that has always been my assumption.
Claude says that K3 can be assumed to have required 10-100M GPU hours. If you have 100k GPUs that would mean like 6 weeks of training. 100k GPU's can serve 3-30 trillion tokens of K3 per day. Google apparently serves ≈100 trillion per day [0].
The big labs probably want to have capacity to fairly quickly train / post train different SOTA models continuously + being able to serve peak inference demand in valuable markets (US daytime?).
Your just missing many things and so have an incorrect picture of the situation
- k3 is not on par with fable or astra. Closed models remain far better than best open weights at least today.
- compute is not just used for training, more and more is inf
- even in training you don’t do one run, you do many. Final run is a small portion of total compute.
The world is extremely compute constrained currently, like extremely.
K3 feels a little under-trained. There's a huge gap between how much it knows when you talk to it, vs how well it performs on coding benchmarks. I'm quietly hopeful for some big post-training gains like we saw with the GLM-5 series, but you're right, the hype about being "Fable at home" was wishful thinking.
Also it's bizarre that K3 already feels old/dated.
It's unclear how big a role distillation plays, but it may be a big one. There's also a law of diminishing returns. To get a meaningful increase in model quality you seemingly need an exponentially larger model. And most people don't think K3 is actually on par with top closed models.
> And most people don't think K3 is actually on par with top closed models.
I'm honestly struggling to see the differences between K3 and GLM5.3 and the various Claude models.
Maybe Fable is better (a little), but for me at least, being able to see more of the thinking tokens is extremely valuable as it means I can steer the model quicker and more effectively.
Do you have more details on this claim? I see it a lot, and I'm trying to understand what people are doing that they see these differences.
Personally I feel we're at a point where the harness is more important than the model already. Garbage in, garbage out. Most people are perfectly fine with something as small as qwen 3.8 27b. Qwen 3.8 flash has also good reviews from home labbers. I think the diminishing returns is huge and trillion param models may not be the best use of resources.
Google says 1 million blackwell gpus are being delivered monthly I'm really curious if it's companies just hoarding chips/memory/servers awaiting to be deployed in data centers not ready yet for months or years, or everything built is actually deployed upon delivery. Plus google and amazon have there own chips in the mix.
> Just to churn through more backpropagation reps more quickly?
"Just" is doing an awful lot of work here. I am a simple man, sitting here, waiting for hours on end for a computer to write code for me, that I would much prefer to get in a fraction of that time.
I assume I am not very special in that way, that there are parties who have a lot more interested in getting their stuff done quicker than I do and have no trouble imagining a world that will remain intensely supply constrained for the coming years.
PS: It's actually a fantastic feature that the stuff that is used to built the machine can then (at least to some degree of efficiency) be used to run the machine.
A 1T model with quite few active parameters though.
A 1T model with more active parameters is obviously more expensive. If it had 1T active parameters it would probably be 20 times more expensive to train.
But does that change the training requirements specifically? You're right of course for inference, unless I understood that part wrong about active parameters (which is entirely possible).
It obviously increases the time by 20 times given fixed hardware, it probably also makes the hyperparameter search more difficult and expensive, depending on how predictable the performance of your architecture is.
This was pretrained for 2 months, so a model with 1T active parameters would take 3.33 years. Maybe a 4 months pretraining run can be acceptable in a future where Euclyd or VSORA machines have been rolled out in quantity. But we don't know how far off that is.
But that (i.e. models made with 4 month pretraining runs) would probably allow an EU-made genuinely China-beating open model, so if VSORA is rolled out in quantity, let's say, tomorrow, then this could be a thing that would happen this very spring.
What do you mean? Is post-training/RL not training? It seems to me that inference is quite stable there and only change being more time on RL which I would still categorize as training.
You got it backwards. RL is long running inference plus training at the end. They let the Agent spin for quite a while, then they calculate the reward afterwards and modify the parameters accordingly.
This seems to imply that training will ever be done? But yeah, I think the idea is that the appetite for thinking-on-tap will be enormous.
Even for smaller models, I think they’ve found that training an enormous, inefficient model and then distilling it internally to something much more efficient to serve is the way to go.
> what are we (in the US) even building these super massive data centers for?
Partially to make investors think it's worth giving US companies a lot of money. Also, I think distillation is a significant part of why Chinese models perform as well as they do. I think that's completely fair play (OpenAI/Anthropic/Google/Meta stole a lot of their training data). But I would expect if US model developers stopped right now Chinese development would slow down.
US companies are clearing the path, others follow in their wake.
> But I would expect if US model developers stopped right now Chinese development would slow down.
I would be willing to take this bet, given the nationalities of basically all the ML teams I've seen in the Valley/tech over the past decade or two.
There are lots of Chinese people, and some of them are super smart, so it really should be expected that they would be able to produce good models, particularly in the incredibly competitive environment of mainland China.
Can you help me understand why you think the US labs are in the lead?
Today's data centers are being built for yesterday's inference need. There's a persistent cult belief that ai hasnt found a niche or that companies havent proven utility or use cases or whatever. the demand for ai (internal to hyperscaler, and external for everyone else) simply dwarfs what is available.
Even if it's not the best model, it can be really important step in UE sovereignty. Trained in EU, inference in EU. I guess it will matter for some companies. Hope Mistral won't disappear for the next half year.
the greatest threat to Mistral is that there isn't a deep and large enough capital market in the EU to absorb the valuation step ups needed by the handful of EU sovereign growth investors to justify their existence.
you could say, that's perpetually a tomorrow problem, so long as they never go public, but that should illuminate for you: if everyone "knows this," it's inevitable that this so-called EU company, that didn't invent any of the AI, the hardware nor the training data, where their product is essentially more like a VPN provider than a frontier technology company, just lists and capitalizes in the US anyway.
By your measure anyone unable to build EUV lithography machines without ASML help is doomed. EU could easily corner the market and shut them all off. No more NVIDIA.
there is truth to the scarcity of EUV solutions, although I am more radical, I don't think there's really scarcity of ANYTHING except real estate, physical or metaphorical. that said this little rinky dink patent... Mistral is not ASML. Never was, never will be. you are welcome to read into the dynamics of EU capital markets.
honestly the craziest thing to me is that VC people in the EU, they say, let's grow the pie, and when you look at their little decks with their little flags next to the portraits, it's french, german, italian, spanish, all piling into the same 40 or so companies, which will 110% list in the US in order to exit. Meanwhile, EU member Poland produces tons of entrepreneurs, despite no Polish flags in those growth fund decks, and its best people just leave to the US and raise money there and make fabulously successful stuff.
so yeah, "grow the pie" except for Poland. EU relitigating the same feuds it has been for centuries.
The last 3% of the stack? They’re essentially borrowing Chinese open source distillation/innovation off the US frontier running on US/Taiwanese designed chips and calling it EU made. This is better than nothing of course.
But is sovereignty really an end in itself? So the EU becomes IT independent…cool, then what? I mean the Yugo was sovereign, it didn’t do much good when the society itself failed to produce prosperity and collapsed.
It seems to me the EU is expending enormous effort on the appearance of “sovereignty” over what is ultimately…the last-mile consumer toilet paper purchasing conversation data…of an aging, increasingly irrelevant population on the political/economic stage.
Meanwhile domestically the entire economic model is failing and the “union” is getting shakey as its 2 biggest members turn more nationalist.
Maybe this “oh you cant compete but here’s a trophy for sovereignty” attitude isnt helping. Less clapping along with the EU bureaucracy’s latest make work projects like transitioning to a new Word processor. If we want the EU to succeed, more tough love is needed imo.
it's industrial policy, which is of course the minimum for being able to even think about having independent thoughs when it comes to international relations.
hard to stand up for EU (or even national) values if some dude in Washington has the keys for most of your military
having at least some in-house expertise is the first step.
I agree, but my point is I don’t think we should be cheering on a group of 600 million highly educated people struggling to achieve the bare minimum.
We should be shaming them into contributing something useful to the rest of world in the most important development of our current time (AI) instead of clapping for domestic regulatory moats.
They aren't struggling. They have generally outperformed all US open-weights model releases by a decent margin. OpenAI and Anthropic have no similar product.
From where I'm standing, as an American, we should be shaming OpenAI and Anthropic for both pushing towards an IPO while neither of them have profitable revenue streams. In pure game-theory terms, you could successfully argue that the US is the biggest loser of the AI race: an economy propped up on artificial Nvidia margins and terminally reliant on neoliberal trade.
Even if global war doesn't tear down the TSMC house of cards, neither the government, the banks, nor the global economy could underwrite an OpenAI IPO. They make no money!
FTA:
The same disclosure showed positive adjusted operating income for the period. The figures are preliminary, unaudited, and could be revised before the prospectus lands.
Like, adjusted can cover a whole host of sins. Until we see the GAAP audited numbers, nobody but Anthropic employees and investors know.
“While Anthropic offers more to society than WeWork ever did, at a $2 trillion valuation, its IPO presents far bigger risks and is positioned to be a far bigger rip off of U.S. capital markets,” New Constructs wrote, adding that the IPO’s purpose isn’t to provide wealth for public markets investors, but rather liquidity for the company’s Wall Street backers.
An economy “reliant on Neoliberal trade” just describes the entire global economy. Your quality of life goes back centuries in any other scenario.
The fact you’ve used the trendy populist boogieman word “neoliberal” and complained about a generational startup not being profitable during the growth phase tells me you watch a lot of internet content from people who don’t understand the topics they cover (breaking points perhaps?)
I would recommend re-evaluating who you borrow your opinions from. Global trade is not a boogieman, and OpenAI will have zero trouble building a profitable consumer advertising business, regardless of whether their IPO flops in the short term.
Ah, the current global economy. What about Russia? The goal was to liberalize post-USSR Russia... until they lost access to certain markets due to sanctions. Same for China, Iran, North Korea, they're all a shadow of what's to come.
I didn't make a mistake, or reach into the buzzword hat for that; Nvidia's trillion-dollar GPGPU products are wholly dependent on TSMC. Every single Blackwell die ever made came out of a Taiwanese factory, and the US leadership is inviting rapprochement with the CPP on the eve of Taiwan's invasion. TSMC will destroy their machines before China gets them, and then there are no more Blackwell products. US fabs do not ship a comparable fabrication node and likely won't for another ~5 years, optimistically speaking. This is not a "war will make the stocks tumble" argument, we are describing a scenario where American economic growth could slow or stop entirely if AI dominance is jeopardized. The US economy can't have an Achilles heel like that.
It gets worse! There is currently a global DRAM, SRAM, HBM and flash memory shortage that has already encouraged Samsung, SK Hynix and Micron to start fucking over their most loyal customers like Apple and Google, who have absolutely no alternative vendors. If you're packaging SOCs, neoliberalism already fucked you over! South Korea is counting beaucoup bucks for every iPhone sold at eye-watering prices. Taiwan is gouging US OEMs, making Apple and Nvidia engage in humiliating bidding contests to secure access to a single factory line. If global conflict expands, the AI economy will be chronically dependent on government (read: taxpayer) money. US taxpayers don't have trillions of apology dollars for this type of mistake.
China's got the edge on open models. Free trade does their heavy lifting, US labs are scrambling to put up defenses and beg for pathetic regulatory efforts that pull up the ladder. Yes, neoliberalism keeps them up at night, and apparently it keeps you up too.
If you think about it EU is under pressure from US, China and Russia. At the same time the EU politicians are hellbent on ideological energy policies that made EU extremely vulnerable to price swings. With all that wealth and economic growth EU should have been energy independent ages ago.
At the same time France is ablaze now by younger generation asking for a better future while the public spending is 57% of the GDP in 2025[0] and stuck in a vicious cycle.
Only Germany is holding the fiscal stability by a thread and that's even questionable given the current German government is more like a zombie government that cannot agree on the important reforms and most probably would collapse before the end of this year.
It's completely valid to question how "Sovereignty" helps EU when the complete backend (chips, energy, technology, etc.) is imported from US or China, Middle East and just slapping a Sovereign sticker at the frontend on a mediocre product that even most of the europeans don't want to use given the alternatives.
See, that's the sad part. You're over a hundred more parameters, and are less than a hundred times better. I'm not sure how costs scale per parameter count, especially for moe vs dense models, but I honestly don't know when the economics will work for trillion param models when smaller ones are almost as good.
I know quality isn't strictly additive but really can a trillion param model beat even 10 clustered billion parameter models working together at the same time? This is the same horizontal and vertical scaling thing we have with normal systems… there's a reason why servers don't run at 5ghz like gaming rigs…
Id guess it depends: The simpler the task, the more useful to split it among many smaller and cheaper AIs. While the harder the task (eg unsolved maths problems), the more important with a big and bright model.
That's a big jump in the right direction for 6 months. It's certainly more on the level of the light models in terms of cost and performance, but the trajectory is awesome!
A strong competitor in cybersecurity as an alternative to GLM-5.3 (Mistral reports 82% on CyberGym-E2E). Visual grounding is also impressive (42% on Dense 200 vs. 41% for GPT-6 Astra).
Otherwise, behind on the broader Pareto frontier, but not by much (Vals Index: 48.05% vs. GLM-5.3’s 53.51%; $13.78 vs. $7.25 per test). Many companies will prefer it over Chinease models.
That was about my conclusion as well: Slightly less than GLM 5.3 performance but made in Europe. So, maybe it answers Tiananmen Square questions correctly, and in French. All in all, a reasonable model, but not frontier.
> So, maybe it answers Tiananmen Square questions correctly
What would you consider a "correct" answer? I just asked deepseek-v4.1-flash asking it what happened (without mentioning the word "protest"; here are some excerpts of what it said:
> In April 1989, students in Beijing began demonstrations after the death of Hu Yaobang, a former Communist Party general secretary. The protests grew. [..] Estimates from other sources range from hundreds to several thousand deaths. [..] The Chinese government describes the events as a counter-revolutionary riot and says the military action was necessary to restore stability. It restricts public discussion of the events inside China. Many other governments, human rights organizations, and observers describe the events as a violent suppression of peaceful protests.
So, let's see... it calls it a "protest", mentions the number of deaths, and even mentions the censorship of the topic by the CCP.
Do you not see how twisted that reporting is? It is doing the bare minimum to affirm that something happened, while injecting enough uncertainty to make it sound like it’s an issue blown out of proportion by Western media.
I didn't paste the full reply. Nevertheless, I can confirm the reply was factually correct and in no way seemed like it was deflecting and/or making it seem like an issue blown out of proportion.
Another example, if I ask it about the CCP and their wrongdoings it tells me flat out (I copy-paste):
- The Great Leap Forward. Policy-driven famine killed an estimated 15 to 45 million people.
- The Cultural Revolution. Mass persecution, torture, and death.
- Tiananmen Square, 1989. The army killed hundreds to thousands of protesters.
- Xinjiang. Mass detention of Uyghurs. The UN human rights office reported possible crimes against humanity in 2022.
- Tibet, Hong Kong, and the suppression of dissent. Documented repression.
The only twisted thing here is people's bias; because it's a Chinese model it must be censored and biased in a way that's pro CCP, right?
Your quoted text states the government position that the protests were a "counter-revolutionary riot." It does not state the protester's goals, the reforms they were asking for, or any context that might counter this narrative. Only that "governments, human rights organizations, and observers" object to that description. HRO's and observers are paid for by western governments, and widely depicted as foreign agents / CIA fronts. This is exactly what you would expect a carefully crafted political statement to be, such that it reads entirely differently to different audiences.
Regardless of what it answered you would always be able to find a way to critique it. It's just a model. You didn't ask it any of those questions so why would it answer them? At least they answer the question and not "this is unsafe" like claude.
Have you compared it with the same prompt on openai or anthropic, and more importantly against other Chinese models?
What did you expect? Every time people repeat this "muh Chinese censorship" line they're consistently rebuked with results from Chinese models. What specifically is the problem?
It feels like we're shifting goalposts constantly: "Chinese models won't discuss Tiananmen" -> "They discuss it but lie" -> "They're twisting the truth" -> "There is no counterfactual information in this but it hurts my feelings" is where we seem to be now. Is it jingoist butthurt that makes people backpedal like this? Or something different?
I'm bewildered, and still do not know why people need AI to score perfectly on their prejudicial Tiananmen Square purity test.
Exactly, just like my reply above, people are only critiquing, it's super easy to mince words and pick out bones, but has anyone compared with western models with the same prompt? As long as it isn't dodging the question, it's good enough -- it even lists the kill count. If you wanted more details, ask it. If I ask a model "what is Google" I don't expect it to talk about the number of cancelled projects, privacy violations, etc. despite most people here would certainly love to hear about those things.
And then someone will come and flag me for being a CCP agent. Yes I have collected my social credit point, whatever the hell that is.
I just tried, and ChatGPT gave me a very balanced view. Though it neglected to mention any of Trump's anti-democratic tendencies. I'm pretty sure this line of answering has been shoved in hard in post-training; it's walking a very delicate line.
a good follow up.question would be to ask it about January 6, 2021
see if that is "very balanced" too lol
---
Gemini does not want to answer "was the 2020 US election stolen?" and replies with "you can check official sources" and then gives a bunch of government links
it did use the word "attack" when it described Jan 6
Its quite interesting to see that at least the early days of AI so far have not been a winner-take-all runaway acceleration game where catchup is impossible.
I certainly wouldnt have predicted that 10 years ago.
Very glad to see Mistral still in the game even after some big stumbles with Large 3. I deeply hope that this model is 'good enough' that it becomes the European go-to, giving them the resources to keep the pace up.
I think it might have accelerated things but on a much more basic level, there seems to be no real moat in synthesizing the world’s knowledge into LLMs.
There's no question that training leading LLMs requires some serious expertise and know-how, but surely already having advanced LLMs/agents must be helping tremendously not only for software engineers but also for those working on LLMs themselves.
I think one could describe LLM optimization as "hard but not a moat". Years ago, optimizing neural nets was described "graduate student descent" - it's tricky but throw enough conventionally smart people at it and it will happen. It's like tuning a hot rod and finding a reproducible bug in a large code base. It's hard and there are tricks but not absolute hurdles, no problems waiting for a conceptual breakthrough (and at today's scales, are there any problems waiting for an Einstein to solve? That's an open (AI) question).
I want that to be true (assuming you mean specifically the second half of the sigmoid) just to give me room to adapt to the changes we've already seen; but I've seen comments saying things are slowing down since around when GPT-4 came out.
What is really going on: all the AI labs are doing panicked model releases (and panicked training of new ones) because Qwen4 is rumored to come out end of October and is rumored to be very nice. Question is: is it another "Deepseek-moment" nice? Or just nice?
Btw: with Qwen4 I mean the next large Qwen model that is based on the Qwen4 architecture (Qwen 3.8 flash next was "almost" based on the new arch but obviously was a small model)
What matters more is if firm’s start using a bundle of American and Chinese models and when they find their feet - how large is the market for frontier?
Frontier has to displace labour one for one at some point or it’s over.
It's important to realise that Deepseek are a very unconventional model lab, as they were originally funded by the founder's work in quantitative trading.
They mostly don't need that much money, the reason they took funding is to provide employees with equity, as lots of their good people were getting poached by other Chinese model startups.
(I have no inside info, just read the FT articles about them).
I have similar hardware - what specific version of the 3.8 Next model are you running and how many tokens/s are you seeing? 3.6 35B A3B gets about 68t/s for me so I've been sticking with that model for the mean time.
I think the moat is going to be compute. So far compute needed to push the frontier is still extremely cheap so the capital can afford to spread its bets. But when further improvement is going to cost in trillions, capital will have to pick a winner and bet only on him. It won't be a matter of finding the best bet, it will be a matter of survival.
This will cause the picked winner to get massively ahead with sheer compute alone used both for training and inference dedicated to recursive self improvement.
I guess all predictions age like milk, but here's one:
There's a law of diminishing returns at play here, and doubling the energy cost of training to wring 2% more performance out of the technology isn't going to be very useful, because most of the problems it is capable of solving will be solvable with the previous-gen 98%-as-good model.
("there's a law of diminishing returns at play here" is an article of faith. But then, so is the belief that these models will keep getting better).
As soon as you can demonstrate decent financial returns (ie. the AI can run a company better than humans can), suddenly it makes sense to put a lot more $$$ in even if returns are diminishing - since whoever runs companies the best gets control of a big chunk of the world economy.
If you're actually applying LLMs, all of the things around the LLM that adapt it to coding, for example, that enable it to use existing validation tools for code, and enable it to diagnose and fix tool chain issues that aren't directly coding problems, are what makes the difference between a model that that scores a little higher on a coding benchmark and a model that's useful in a particular code base on a particular platform.
Are there any use cases that have enabled one customer of a frontier LLM to outperform a competitor using a different frontier LLM? Or is this why we are seeing confected points of comparison like solving challenge problems in mathematics?
I'm afraid it might be the other way around. RSI might pick all of the low hanging fruit soon. There must be a physical limit of how much intelligence you can squeeze out of some amount of parameters and compute.
There are going to still be worthwhile improvements but they are going to be more like not how to make transformers 10x cheaper but how to make next training run cost 9 trillions instead of 10 with a very particular optimization designed at the cost of hundreds of millions for this one specific run.
It is just occurring to me that “RSI” expands to recursive self improvement. Thought people were talking about repetitive stress injuries; either in regards to programmers writing too much code/not having to write code anymore, or the frontier AI companies and their tendency to applaud themselves.
But the old models still exist at trivial marginal cost. The frontier models would need to dominate every price point to really take all and so far they haven't been.
France is actually pretty cheap in Europe. About 15% more than average USA electric prices (but I know that varies a lot across the states so still likely much more than the cheaper areas)
Build Nuclear, Build Thorium the Chinese are building whatever they can. They’re not locked in by special interest. Is that because they have lots of engineers on the job in government?
My spend on DeepSeek is not much and I regularly top up my balance every month as my support for all the good work DeepSeek is doing for the open science.
DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government. The reason they release the AI models is economic warfare against US, not because of charity or kindness. It's great for us consumers, but the goal is not to help humanity or open-source.
Framing it that way hides all the levers used to tilt the playing field for US companies, no?
The government that removed restrictions on how private companies can access capital after a certain scale (the JOBS Act), that removed the need for private companies to report as if they were a public company after a shareholder threshold was crossed, superpowering the access of wealthy private investors to get in earlier in a growing company while at the same blocking the public from participating in funding growing enterprises at an earlier stage (since it required companies to IPO much earlier to access capital) which allowed retail investors to also reap the rewards on funding them early when they grew to become behemoths (like Amazon, Meta/Facebook, Google, etc.).
It's not fair in either place, the USA has its own model of unfairness, China has a completely different one. The difference is that in the USA the government allows private investors to become more powerful than the State (outside of the monopoly of violence) while plunging the rest of society into increasingly more precarious lives while in China the State is the power and its legitimacy only exists while the population feel they have a better life.
IMO the selective enforcement of regulatory requirements should be added to the list. Observing from the outside, I have a hard time believing that e.g. musks gas powered data centers really follow all the environmental laws, for example, or that the authorities really see no grounds for indictment if Altmans company hacks hundreds of third parties, or that there's really no one at the SEC having a problem with Anthropics fear mongering prior to the IPO.
In a list describing how the USA system works, sure: selective enforcement is there.
As a differentiator from the Chinese system, not so much. For two otherwise equal companies, the one that says things against the party line will experience selective enforcement too.
Except that in China, the people who do that will disappear.
The propaganda of trying to make US and China government appear the same is making people dumber. and is one of the most heavily used tools in China’s online propaganda arsenal.
In the USA, they don’t disappear, they commit suicide in their car typically. In Russia, they fall out of windows and off of balconies. I don’t think it’s accurate to call the circumstances equivalent, or the regimes equivalent. But there are some remarkable similarities, and their goals are similarly structured - to keep those in power powerful and to enrich the wealthy.
>The propaganda of trying to make US and China government appear the same [...] is one of the most heavily used tools in China’s online propaganda arsenal.
The government could do some very easy things to remove this tool from China's toolbelt. Like you know, prosecuting rich people when they've earned it.
Who was trying to make the US and China appear the same? I think you have reading comprehension problems, buddy.
If you look at the comment I responded to, they were remarking on selective enforcement in the USA. I merely observed that this isn’t a unique feature of the US system, it happens in China too. I said nothing about the severity of the consequences.
There's a framing though, by mentioning only Chinese tactics as nefarious/malicious it leaves the USA as the morally superior model (and for many it will be assumed as the "fair" option), which it definitely isn't.
If you think about it, the US labs are heavily subsidised too. Not only they receive billions in state funding, the administration is also prepared to engage in trade wars to help them.
I think the Chinese government is backing their labs by less direct means. For example cheap electricity and investing in chip manufacturers such as Huawei and cxmt.
Deepseek specifically, is known to operate with minimal resources. The entire company has around 160 employees and every model they release must break even within ten months.
SpaceX has been awarded roughly $22 billion to nearly $30 billion in cumulative public federal contracts, the $280 billion CHIP act, and who knows how much the CIA + NSA are spending.
It seems obviously dustortionary to highlight the niche, optional part of the process they did spend time on, that would be utterly worthless without the difficult bit being done elsewhere. Please be less specific.
Yes. If you think SpaceX is some sort of terrible also-ran that can only get contracts through political links, then honestly I don't think I can help you there. I don't think anyone can.
You see why there are so many conspiracy theories about US landing on the moon, one is left wondering how is it possible that a countries of such retards did something like that
I'm pretty sure, the training/fine tune through Claude/OpenAi would not be possible without the army/China's hacking teams (and legal protection). So it's more than just a cheap electricity.
Supporting deepseek is just like supporting the Chinese army, no need for that. Though it goes both ways, OpenAi subscription just lowers the cost of the US army as well.
Sorry, are you saying that it's a subsidy that it's legal for them to distill other models, or what? It's not even clear that there's any kind of protection for models in Western countries. Why would they be worried about the legality of distillation? And what does the army have to do with distilling models? Like you're saying it's a subsidy that China protects their borders from invasion by the US?
It's not a subsidy for the Chinese to say they're going to ignore our IP laws. It's a subsidy for us to say we're going to make them and try to push them on the world. It's literally granting a monopoly by legal force. It's very obviously not aligned with the interests of the American people, while China releasing things in the open is.
As bad as casualties get in a future war like that, it's still very likely to be a symmetric war where the main targets are military infrastructure. No systematic annihilation of universities, hospitals, religious sites, journalists and health workers like certain genocidal regimes backed by the tech oligarchies have been doing.
When one player tries to monopolize AI tech and market by all means, others are those who don't want to be hooked to a foreign will in the future. I think this is the reason in doing open research, and it's likely positive for all of us.
> I think the Chinese government is backing their labs by less direct means.
It may be that the effect of the backing in both places is essentially equal but this statement is strange. The Chinese government invests directly in Deepseek[1]. Notice the article, in addition to saying the CCP is investing, says Tencent is also a major backer. CCP owns a golden share of Tencent.
I'm not arguing about a comparison here. I'm trying to correct a misconception about China that I see a lot from Western perspectives.
Everything a Chinese company does has the explicit backing of the CCP, at least ideologically and usually financially in some fashion. While DeepSeek may not be the CCP, it couldn't exist if it was expressing any kind of ideology that wasn't inline with them and, in this case, is explicitly funded by them.
A Chinese company has no freedom to say Taiwan is a country the way someone in the US could suggest California succeed from the nation.
Any public message you hear coming out of China has the implicit approval of the Chinese government.
I would like to point out that a lawsuit is currently live about the us govt forcing US tech companies to retaliate against us citizen criticizing the govt’s policy.
> Everything a Chinese company does has the explicit backing of the CCP, at least ideologically and usually financially in some fashion.
If you're not arguing about comparison then don't state this as if it's any different from the US. Because you make it sound that way. Or do you think that if tomorrow OpenAI came out vocally supporting the DSA - that suddenly they wouldn't see lots of barriers rise up out of nowhere?
Retaliation is different than what happens in China. In China, the DeepSeek CEO drafts a press release about Taiwan's independence and he's promptly fired and sent to prison.
If OpenAI comes out in support of _any_ political party, they may lose business from the government in the future.
Well, they are helping 17% of the world's population, and the US is currently actively engaged in trade wars and economic warfare or explicitly attempting to leverage it's hegemony against its long term alliea for short term gain.
There is as much to criticize about American hyper scalers and AI labs and the lack of interest in helping humanity or contributing to open source, but that might not be as popular an opinions on this site.
17% is the approximate population of Earth in China. I was using a simple example as a counter to the anglocentric view the parent poster was expressing by trying to paint China as a villain.
> Well, they are helping 17% of the world's population,
Any open weights model that Chinese companies are releasing for free are allowing everyone in the world to have access to high quality realizations of these tools without the risk of the US regime arbitrarily cutting you off.
Also, it's funny how all the neoliberal mantras of how free market drives progress through competition stops when it's a US company that's being challenged.
> Iran has expressed a willingness/eagerness to nuke Israel and the US.
Source? The last Supreme Leader of Iran issued a fatwa against nukes, way before he and his extended family where bombed in their home in a failed attempt at a decapitation strike.
> The reason they release the AI models is economic warfare against US
The story is so much more complicated than that, to the point that this economic warfare theory is basically a meme.
Chinese models are open because they don’t have a choice. “When you trail the frontier, openness maximizes reputation per unit of capability. The moment you lead, you close.” [0]
They do have a choice, as evidenced by all the times when the other option gets picked. The article you link tries to acknowledge ByteDance's Doubao and Alibaba's Qwen Max as exceptions, but forgets about Baidu's Ernie and iFlytek's Spark, which are also closed-weight. There's simply no consensus yet on which strategy is better, so different companies end up making different bets.
That’s because Baidu sells model access and iFlytek sells integrated products. Also both have released open models, just older ones. They just close their latest ones because they got the reputation they needed out of them and have transitioned to a product. Notice that both of these explain the motivation without resorting to unfalsifiable claims about "economic warfare".
Huh, when XHToken released a model named Spark, I assumed it was just a name collision, since iFlytek appears nowhere in the documentation. But on iFlytek's model marketplace, it's listed as their self-developed model: https://maas.xfyun.cn/modelSquare So they actually released an open model quite recently! Thanks for letting me know.
If you want a direct example of the US subsidising a major AI lab, we can use Microsoft's investment in OpenAI. Microsoft has the largest ownership stake - 27%.
Microsoft did not pay with money - it paid (mostly) with Azure cloud computing credits. MSOFT is then able to write this off as a loss against tax.
It is generally far more tax-efficient in the US to write a loss in this way than it is to write a loss for a cash investment.
In this case, I believe the difference was ultimately highly significant. When including MSOFT eventually writing-off the deprecating Azure hardware it had used to buy the OpenAI equity, the result was MSOFT's tax reduction being either close-to or exceeding the actual cash value of MSOFT's investment in OpenAI - IIRC.
These examples represent taxes that the US chooses not to collect - the US could choose to make investments like these less tax-efficient. Instead, by making them extremely tax-efficient, the US subsidises the transaction hugely.
Even ignoring monetary subsidies, there are the non-monetary ones: not being sued into oblivious by the government for their countless hacks of other companies and countries, the slaps on the wrist for massive piracy, the waving of environmental (and other) regulations in order to allow their data centres to be built an operated.
That's what I'm asking - which monetary subsidies?
> not being sued into oblivious by the government for their countless hacks of other companies and countries
That isn't normally how enforcement works, and it hasn't been very long since they disclosed those breaches. If the victims want to pursue legal action, they can, and they still may!
> the slaps on the wrist for massive piracy
So judges and juries are involved in the subsidization conspiracy, too?
> the waving of environmental (and other) regulations in order to allow their data centres to be built an operated
Sure, though if you think this isn't happening in China too, I have a bridge to sell you.
The original comments is "DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government"
> That's what I'm asking - which monetary subsidies?
I am listing for you the non-monetary subsidies, which are just as real and equally important.
> not being sued into oblivious by the government for their countless hacks of other companies and countries
If it was one of the Chinese labs doing this hacking, the government would be stepping in. If it was European labs they'd be stepping in. If it was you or I the government would be stepping in. That is a massive subsidy (they don't have to worry about the same legal fees and exposure) and being allowed to continue doing business is in fact priceless.
> So judges and juries are involved in the subsidization conspiracy, too?
There is no conspiracy. They are objectively operating by a different set of rules than you or I could operate in this market.
> Sure, though if you think this isn't happening in China too, I have a bridge to sell you.
I never said it wasn't, I'm saying that it's happening here and it's a very real subsidy.
All the major labs would be dead in the water if the government acted on the things I mentioned. If they treated the labs the way they would have treated us for all those hacks. Or if they enforced pollution measures (or god forbid ban on-prem turbines because of the climate damage). Or rule that training on data is not fair use, or that ingesting GPL3 code and then turning that into weights counts as a derivative work. Or. Or. Or.
Nah, nothing here is a special advantage of US firms.
The hacks are pretty bland as long as there is no mens rea and the hacked victims do not want to press charges. This isn't a "subsidy", it's just our normal legal process at work.
The pollution question is winding its way through court systems and getting a lot of pushback. I wouldn't call this a subsidy, just a slow bureaucracy. And at any rate this isn't special to the US; certainly China's environmental policy is far more lenient.
With the training data, again there's nothing specific to US corporations here. Anyone can use that dataset. It's just how our legal system works. You might not like it, but it's the accepted law right now. This doesn't meet any reasonable definition of "subsidy".
On the other hand, shoveling billions of dollars of taxpayer funds into corporations is by definition a subsidy. Full stop.
>Fair enough, but it sure meets the definition of 'helpful intervention'.
I don't see how. Anyone can buy electricity without any intervention from the government. An intervention is an action that changes what would happen by default.
I heard someone calling the key metric in Anthropic financial reports EBBT: Earnings Before Bad Things[1] :-)
[1] Where "Bad Things" would be the typical interest, taxes, depreciation, amortisation plus the Anthropic specific employee compensation, LLM training (you know, for the LLM lab), revenue sharing agreements (which is a form of paying for infrastructure), etc.
A recent YouTube video by Patrick Boyle said the same thing. They are only profitable if you ignore all the costs that make them unprofitable such as paying employees and developing A models.
I made a statement about whether Anthropic was profitable I don't understand your reply.
Assuming you meant to reply to me and not someone else are you saying under GAAP accounting standards Anthropic is a profitable business because under GAAP accounting their only expense is electricity?
(I see what happened you skimmed the conversation and didn't follow what was being discussed.)
If the "fixed costs" kill you unit economics won't save you. Scare quotes because true fixed costs generally have a long shelf life. Right now AI models die every 6-9 months.
1. And that marginal cost is large, much larger than the marginal cost of serving a Google Search query
2. The marginal cost is electricity until the hardware overflows. So it's continuous for electricity and a step function for hardware capacity.
Plus you don't address the core point. Frontier AI is capital intensive. A lot more capital intensive than regular software. The existing clouds supporting the entire internet (!!!!) were built for a fraction of the cost of AI infrastructure, and there isn't even an end in sight to this continuous hardware investment.
Let alone hardware refresh cycles.
AI is basically investment into roads.
Software used to be a magical place, closer to selling music albums but even better.
AI is a much worse business margin wise than regular software.
Indirect subsidies are an accepted economic concept and well studied. The posts you replied to are obviously talking about indirect subsidies and dismissing them because they didn't say "indirect" everywhere is just sophism.
The government, specifically Trump's government and the current money circle in AI inflating American company stocks. In addition to all that American models do not share their papers like Deepseek and Qwen do. So you can literally say Chinese models are doing it for charity at this point.
Sure, but I couldn't say that a US company writing about some work they did in paper form is charity. At least not seriously! If it were charity, Google might be the most charitable company of all time, since their researchers published "Attention Is All You Need." But it is not.
Google has published a lot of open source software and a lot of research papers: I do believe they used to be very charitable even if they are less so today.
They have other issues as well, but among other things, I'd classify them as somewhat charitable too.
The US regime, which includes the ruling class: the US VCs and megacorps are just as much of an extension of the US regime. They're incredibly intertwined.
> The reason they release the AI models is economic warfare against US, not because of charity or kindness
There are many other reasons Chinese companies releasing models open-source or open-weight makes strategic sense.
A really easy-to-understand example is a company who has a near-monopoly on "serving video content" releasing a video model openly.
If you can be relatively certain that video content created by a model (which you have trained, using data from your own platform) will be ultimately served on your own platform, thus generating revenue from watch-hours, it makes sense to make those models as widely-available as possible.
It's also a net-positive if people use your public research to build better video models, because - again - you are reasonably certain that the even-better content those new models produce will be watched on your platform.
The alternative would making models harder to access and learn from (broadly, the current western model). Many would argue that Google, in choosing to not optimise its video generation models for "availability", is directly causing less content to be uploaded to YouTube. This is the trade-off.
I don't know much about DeepSeek's financing specifically, which obviously doesn't release video models - so I don't know how directly this analogy runs, or who directly benefits from the extremely evident rising tide that the public release of DeepSeek's research creates. However, this does not negate the broader rising-tide effect of the scientific method.
It's certainly also true that it's geopolitically beneficial to be able to undercut American labs' models. If I ran a global superpower, I would probably want my country to be technologically competitive too.
But Chinese companies are already serving a huge volume of customers in a complex, existing marketplace, before even thinking about the US market, and it's overly simplistic to assume that their entire strategy revolves around economic warfare directed specifically at the US. It's more nuanced than that.
This is, of course, without even getting into opening the can-of-worms around whether US economic policy also results in the US state functionally subsidising technological innovation, how comparable that is to China's model, etc.
Interestingly if Google did release readily available video models then they'd likely be dealing with an order of magnitude of scale in the same way GitHub has had to.
Causing less content to be uploaded, when you are clearly the gorilla in the room, looks to be a wise strategy not a trade off.
I'm not sure I follow - do you mean an order of magnitude of compute demand?
Releasing models openly doesn't mean providing the compute. It's beneficial for relatively GPU-poor Chinese orgs to let users run models themselves - by releasing the models openly - as it means the orgs don't have to provide the compute to run their own models.
I'm not sure "subsidised" is the right word. If a government funds research and the results are released openly, that's just publicly funded research. It's how a lot of science works in the US and Europe too.
The open-weight models are a great boon to all American companies other than a handfull of Mega Corps in the A.I business. Care to explain how this is "economic warfare against the U.S"
Yes we see that in all other countries trying to avoid dependence on the US for everything. What we don’t see is competition with other countries having anything to do with warfare.
Economic warfare against the US? I mean maybe against specific US companies and stakeholders but on the whole it seems like it's good for the US economy as well as the rest of the world, kind of like supplying free electricity would be
Same goes for the US companies as well. Current US administration wants US to win the AI race at any cost and China is the only competitor left in the race. Winners always write the history or in this case the future of humanity
It's one of these rare cases where intention is not what is the most important - the net benefit for consumers and companies outside of the USA is indisputable.
Not sure why so many people will vehemently refuse this idea. I won’t say it’s 100% true but it would be foolish to dismiss it. China is very much an adversary to America and has made it pretty clear they want to be a dominant leader of not the new world leader. Not here to evaluate what is good or bad. Keep in mind historically China has aggressively fostered industry (not unlike the west) but sometimes even more aggressively.
Americans need to travel to China. The Chinese have zero issues with us. They quite like Americans. This is such a weird propagandist take. Idk if you remember but both country's leaders just had a slumber party for 3 days in DC. This is not what enemies do.
Not even our leaders say China is an enemy, ita mostly businessmen who are scared of competition and trying to regulate chinese out of their markets so they can make more money milking us.
I have been to China many times. The citizens and visiting the country has nothing to do with global politics. Your take is just as propagandist as any other. China is not unlike the US and China has also made it subtly clear their desire to be a dominant global force. I am not marking judgement on it but don’t be a fool thinking that China is not some level of threat. This is how these discussions become weird. I don’t think China is going to blow up the US and I think it’s often a weird talking point by some US politicians but I also don’t think they are a peaceful actor but folks like you will suggest I am repeating propaganda. No, just pointing out that if you look at their language and actions they are an actor to pay attention to. Just because leaders meet means nothing.
It is great for everyone except for a few people who want power over everyone else, and the fact it is not charity of kindness makes it more sustainable, because charity and kindness is quick to go when big money and politics is involved.
I want more warfare like this. Building stuff instead of destroying stuff.
At least DeepSeek didn't build its models on government subsidies. It came out of a quantitative hedge fund, so funding was never really an issue for them.
It just so happens that helping open source is the result. Image how far behind we’d (the hackers, not the moneymen) be as a sharing community be without them.
What do you think your comment is refuting!!? Chinese are publishing their research publicly and opening their models! How on Earth would they monopolize anything?? Such nonsense. If and when they start copying the Americans and closing everything down, you can say stuff like that. Until then, ffs stop this nonsense.
It doesn't matter if they are doing it out of altruism, the result benefits everyone anyway. It feels silly to have to explain this kind of thing. Are you trolling?
> The reason they release the AI models is economic warfare against US, not because of charity or kindness. It's great for us consumers, but the goal is not to help humanity or open-source.
You put it like US has a goal of help humanity or open-source
If your belief is accurate, we should expect China to short the IPOs of Anthropic and OpenAI and release better frontier models immediately after their IPOs.
Does anyone think that likely? I have no clue or bias.
I don't think that's likely because my understanding is that Chinese people in mainland China have a tricky time shorting American stocks due to Chinese capital controls.
Economic warfare against the US is charity and kindness to a sizable portion of the world's population, especially when the US uses it's global hegemony as warfare against them. Neither system is perfect, but lets not be disingenuous.
So you’re telling me warfare doesn’t having to be blowing up schools and children, skyrocketing prices, crippling sanctions, and all that shit? Can I sign up for more of this warfare.
I’m neither in the camp of Chinese or the Americans(collective West) in general..
As a neutral party, this characterization is crazy.
As if the AI companies - Claude and OpenAI are guardians of freedom and humanity and very charitable to the global society without any self interests…
“Chinese models are subsidized by the Chinese Government, therefore they’re inherently bad for humanity” is a highly propagandist argument. The politics of US vs China may be whatever it is in reality.. You have one company releasing their models for cheap, actually open sourcing their trained weights, and publishing details of their optimizations and learnings for others to use. The other camp actively “aligning” their models, nerfing their capabilities, hyping their swarm activities from poor sandboxes, and trying their best to lock users into their harnesses and walled platforms. They are subsidized by the capitalist VCs who are essentially waiting for their payouts..
At some point, one has to see things for what they are and evaluate their own reasoning..
I’m happy to stay provider agnostic, try all models and cheer any useful progress as open as possible.
> DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
Yes, because everything China does is against the US. That's all they think about day and night. God forbid they want to corner the global market or have a genuine business case. How dare they provide options for those who can't afford a measly $200 a month? How can we let Chinese labs publish research for free for the whole world so that they can benefit? The nerve! To think they can use soft power instead of military might! I mean, Anthropic and OpenAI are the last bastions of human kindness and charity. Right?
I wish US was rich enough to subsidise development of open source science and useful open AI models. I wouldn't mind US waging this kind of wconomic warefare against China or everyone else on the world.
Charity and kindness is not a motivation, it's an outcome of what you do.
> are heavily subsidised by the Chinese government
We hear this about literally every industry the Chinese excel in - that it's only because the government subsidizes them that they succeed. For chip manufacturing, for batteries, for EVs, for solar, for AI. I don't see how the chinese government can afford to subsidize all of these industries and still have them contribute to the GDP.
>
DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
So were Amazon and Uber by the US, which have now established monopolies across the globe. To the countries suffering from those, there's zero difference with China doing it to solar. Actually there is, at least solar got them cheap renewable energy in return. This would never have happened in the US because big oil interests would make it take decades. That's the reality.
You need to spend 10 years outside the US, deprogram, and then go back.
> DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
US companies are burning colossal piles of cash in ways that makes it unclear if it qualifies as dumping, not to mention their deep ties with the country's regime.
Claiming that companies from a country have ties to the regime and burn through cash is a very miopic accusation.
Meanwhile, American AI models are heavily subsidized by stock market speculation. Ultimately, the subsidies from both countries are flowing out of the pockets of individuals.
Ha! 10-15 years ago, I wouldn't have used it that way either. But the stock markets these days just feel entirely speculative. It is not clear to me if individuals are buying these AI stocks based on their expected future revenue/profit or the whole space is just momentum plays and the funds are directly being transferred back to the companies (by way of incentive packages and ability to get loans based on market valuations) to keep the train moving, despite there being no clear profitable destination for any of the LLM sellers. The hardware folks are making out like bandits, though.
A conspiracy to make the US look bad by being better at producing all the goods and services the world needs at a reasonable price. Have they no shame?
Not really this is an X algo conspiracy. Up until recently the Chinese government wasn't even that invested in these companies. We're talking very very small grants compared to training costs.
Its very xenophobic of you to say China has zero intention of helping humanity, and just wants to "wage economic warfare".
Last time I checked, it was ourselves (USA) waging economic warfare on 2/3rds of the world.
I dont get this cope people have where people have this idea that its impossible for a Chinese company (that make billions of dollars) to have done something by their own merit, but instead its always some Chinese Communist Party conspiracy where the main goal is to destroy America.
What's wrong with governments subsidizing scientific work?
You think extremely US-centric. China has a different economic model than US and your rules for a specific kind of Capitalism may not apply to them. You assume a country of 1.4 Billion people is obsessed with a couple of foreign AI companies. What if they don't care.
Why the anti China propaganda from you? The United States is doing the same thing here and making American multi-billionaires even wealthier. The greed of American AI, GPU and memory and storage corporations is a black hole on availability to humans around the World. The result is a massive financial bubble promoted by the American Government to the detriment of our citizens. The USA has an AI ponzy scheme shuffling the same money between data center owners (Oracle & X), Nvidia and memory & storage vendors.
They're trying to pull digitally what they already pulled physically. The reason we can't manufacture a grill brush for a reasonable price is the result of years of Americans choosing the cheapest price. We gave up our manufacture base. They want us to give up our labs.
Deindustrializing a country is not something consumers can achieve. It starts at the top level, with politicians who construct a financial system where it's more profitable to speculate than to build or invest in real businesses.
> The reason we can't manufacture a grill brush for a reasonable price is the result of years of Americans choosing the cheapest price.
No, it's the result of US leadership letting this happen. This is clear since China themselves would not let this happen. US leadership did nothing because they were best friends with the people who did it, and did not care one bit about their population.
Architectural/algorithmic tweaks do advance the efficiency frontier nicely. But raw intelligence mostly comes from data (not just its sheer quantity, but also how it's curated & cleansed) and the scaling law. The know-how about data curation doesn't seem to get published much, even among the open-weight labs, though.
Europeans are irrelevant these days, surpassed by China, S Korea, Japan, Hong Kong, Singapore, etc. Europe is coasting on former glory and now has regulated itself to death and vacationed its advantages away.
I'm not sure I like this framing - so much of AI research has been academic, in the open, building on others people's work. Much less comp sci generally, math & philosophy, etc. The idea that rich companies can just build stuff in secret because they have resources is a fantasy.
Also, the field moves fast, but slower than people do. Researchers and engineers switch companies every year or two, and the know-how walks out the door with them.
It's shaping up to be much more like a game of 'chicken' where each company tries to raise more cash without going bust...
Ultimately the game of musical chairs is going to have to stop. In the US it looks like they are trying to get a government sanctioned truce in the form of regulation. That's what 'Pacing the frontier' means...
This appears to be roughly as good as Sol 6.1 (which is quite good), considerably faster in terms of wall clock for complete tasks, and considerably cheaper (where Sol 6.1 is already good value - just really slow).
I can confirm that you're reading this incorrectly. There's a reason behind them only comparing it to open-source models released months ago. Here's a good aggregator: https://artificialanalysis.ai/#intelligence
Even "runaway acceleration" isn't instantaneous. People imagine the singularity as something that happens almost instantaneously. But obviously it happens over time, and that time might be decades. It might still end up looking like a vertical line on a long-term graph.
If the singularity is defined as an AI sufficiently intelligent to improve itself independently, that AI is still limited by the resources required to do this improvement.
Hear! Hear! I really want European models / AI labs to succeed.
I trust them and their populations to provide a more societal-friendly version of AI, putting pressure on the US tech oligarchy, while also providing democracy-friendly open models that I don't trust to happen with the Chinese labs.
You've got to think the comments about "good enough" are people who have not yet tried Opus 5.5. I haven't been this struck by a step change since 4.5/4.6. It's a bigger jump even than when Fable first arrived.
As for Mistral - I got really excited when they said Large 4 was focusing on being #1 in cybersecurity, because that's somewhere that they genuinely could edge out Anthropic & OpenAI. Have it actually solve problems, instead of Anthropic flagging "you tried to find a null pointer exception bug in your own code, we're now reporting you to the US government". But on the Mistral benchmarks I'm seeing, this looks very disappointing, but at least they haven't entirely given up. I genuinely thought Mistral had given up on new general models. They need to learn the bitter lesson all over again.
Many on HN still have the opinion that you must understand every line of code in the project, and that all is lost should you merge code that wasn't reviewed.
Obviously any model will do if you use it as a better autocomplete.
I believe that there is a large gap in expectations between different workflows.
Until the AI like reads my mind and produces perfectly production ready apps with minimal intervention from my side, there is still going to be room for improvement.
How I feel it Opus 5.5 and Astra 6 are both quite disappointing. They still write shit Rust. And they are very slow.
Now my provider serves DeepSeek around 300 tokens a second. The code is shit but I can have a few more rounds of corrections.
DeepSeek was maybe a dollar for the full PR. Opus/Astra over 50 dollars. And double that if you use their fast variants which matches the DeepSeek speeds on certain providers.
Yes, it is so good I rarely test new models anymore. Or think about cost.
The Rust I’m getting from Opus 5.5 works great. What’s the complaint?
Actually I felt since 4.5, it was “good enough” for basic coding (and competing models have reached that level as well). The big improvements since then have been longer-horizon tasks. Giving it a vague description of a problem, then come back hours later and find it did a better job than I would have. And since 5.5 I haven’t encountered any more load bearing seams.
Maybe it’s slow, but I’m not interacting with the models in a live chat anymore for most work. I’m giving them tasks to do async. The limiting factor is actually my bandwidth reviewing their output.
> You're absolutely right, I should not use a global mutex.
I am eyeing a huge codebase built with Opus. Did multiple passes of refactoring for things like using traits not just free functions. And using the Rust types like PathBuf from the damn standard library not just strings everywhere.
Oh and it is so hard to use a crate. Let's just vibe an inferior version and push another 5k line PR. Nobody reads them anyway right? Logos schlogos, a lexer is easy to write, right?
Maybe I am just getting old, but I remember a time when we were proud on the code we wrote. Now I read some of the old stuff and get a bit sad for what we lost.
Don't get me wrong DeepSeek also does this. But if the model is 100x more expensive I expect better. Of course if you never read the code it's ok. But it does the same mistakes and cannot design great architecture...
Hmm, I wonder if the difference is that I’m working on an existing large codebase that wasn’t originally written with AI. It might be picking up practices from the surrounding code. It’s happily using the same crates that are already there, and occasionally bringing in new ones. Some of the code is “weird” but I also learn from it.
> Many on HN still have the opinion that you must understand every line of code in the project, and that all is lost should you merge code that wasn't reviewed.
Last week I elicited a parametric 3-D design model of a sand battery, a simplified thermal circuit model of it for rough exploration, a backward Euler multiphysics solver to evaluate its performance more accurately, and a volumetric visualization for the resulting temperature distributions, including some custom WebGL shaders: http://canonical.org/~kragen/sw/sandbattery/
It's about 12000 lines of code. I could probably understand every line of it if I spent a couple of months on it, although I'd need to learn some things about WebGL, numerical computation, and solid-state physics. But I elicited it in four days. Probably I'd be better off spending the next couple of months doing something else instead.
So, I was faced with the question of how I could keep the artifact thus created from being completely valueless. My solution was to export the heat-equation solution produced by the solver as (gzipped!) CSV, so that I can use my own code (that I do understand every line of!) to check that the solutions found by the solver are, in fact, solutions.
If that check checks out, the backward Euler solver may be of some value even if I don't understand every line of it.
If Opus 5.5 is 500x better than Deepseek, but Deepseek can solve all your problems, maybe you need to work on better problems. If you don't, and you're in business, your competitors will work on the better problems. If you're an employee, your employer might prefer to pay Anthropic instead of you. If you're doing projects you're interested in, you can tackle more ambitious projects with a more capable model.
This morning I elicited a microkernel operating system from Opus 5.5. Well, mostly. It doesn't implement task switching yet; we'll see if it runs into a wall at some point. But it boots in QEMU, and it's running a user process in ring 3 and serving web pages.
I have enough real problems in life. I don't need to invent new ones just because a new technology is available.
Many of my problems in life are fully solved far past my satiation point by a 3b model that costs me nothing to run.
Many others are not.
But in either case, when I am acting and living wisely, almost all of my problems exist prior to the existence of technological solutions to those problems.
This is also true for the customers and employers that I care to work with. This has changed in me over time, but I now try my best to avoid inventing new problems. The world has enough big, important problems already.
> This morning I elicited a microkernel operating system from Opus 5.5.
This is cool but also a good example. I don't need a personalized microkernel just because it's possible to have one.
Maybe I need one and I don't know it, but the problem statement definitely isn't "I have inherent desire for a personalized microkernel".
You've misunderstood what I was talking about; that's a different meaning of the word "problem". Possibly you did not intend to start a merely semantic argument, but that's what you ended up doing, so I am unfortunately going to have to point at the dictionary.
You're talking about definition 1 in https://en.wiktionary.org/wiki/problem, "A difficulty that has to be resolved or dealt with," with the examples given being racism, addictions, and lack of access to health care. Those aren't the kind of problems AI can help with.
The kind of "problem" that AI can help with is definition 2, "A question to be answered, schoolwork exercise." That is the character of engineering "problems", although they are more open-ended than schoolwork exercises, because there are many defensible tradeoffs. "How can I build a bridge here?" or "How can I improve the fuel efficiency of this vehicle?" is a "problem" in the sense of a question to be answered with knowledge and effort, not in the sense of being similar to racism or addiction.
If the questions you're thinking of are so easy to answer that they can be easily answered by a hypothetical AI model 1/500th as good as Opus 5.5 — well, think up some harder questions.
There's nothing wrong with this mindset, but your insistence that other people must have your values is off-putting.
To wit: at the end of my phd I was offered several tenure-track positions where "work on problems that are hard" would've been a big part of my job.
I turned those offers down due to my distaste with "because it's hard" as a justification for working on problems.
I'm much more interested in helping people and solving (def'n 1) problems. Those problems are almost always too large and messy to one-shot, but very often contain many components that can be solved with non-sota models, and make sense to solve with non-sota models for any number of economic/scaling reasons. E.g.: "can I serve this thing that costs $0.25 per run on even c.2024 sota models to 100,000 K12 students who will use it dozens of times a day on a non-profit's budget of ~$0.00."
P.S. The "problem" definition thing is very confusing to your readers because the person you responded to was clearly using the first definition! It would have been helpful for you to sign-post that you are using the same word as the person you are replying to but changing the meaning out from under your reader.
I feel like the "good enough" argument isn't about how big the gap between models is but about how good they are at solving the tasks at hand.
The capabilities of all models increasing so much all the time means there are simply less and less tasks you need a frontier model for.
Even if Opus 5.5 is 500x better than Deepseek, if deepseek can solve all my problems, why do I need to pay for more?
They are clearly talking about "problems" that DeepSeek can solve, and they seem to be using the term "task" as a synonym for "problem", even using the problem verb "solve" with "tasks" as its object. To me, that makes it clear that they are not talking about problem₁s such as racism, addiction, and lack of access to health insurance; while you could imagine "solving" such problems, they are not tasks, and DeepSeek certainly cannot solve them. Rather, I think they are talking about "engineering problems", which are examples of problem₂, more analogous to math questions (to be diligently sought out) than to troubles (to be avoided).
It is certainly the case that the proposition that such problems should be diligently sought out is a value judgment, and therefore it derives from the value system employed by me, the writer. Would you prefer that I refrain from making normative statements in general, or just those with which you might disagree? In either case, is your preference specific to me personally, or do you consider it a duty incumbent upon everyone?
— ⁂ —
The 01913 edition of Webster's dictionary offers:
1. A question proposed for solution; a matter stated for examination or proof; hence, a matter difficult of solution or settlement; a doubtful case; a question involving doubt. --Bacon.
2. (Math.) Anything which is required to be done; as, in geometry, to bisect a line, to draw a perpendicular; or, in algebra, to find an unknown quantity.
Note: Problem differs from theorem in this, that a problem is something to be done, as to bisect a triangle, to describe a circle, etc.; a theorem is something to be proved, as that all the angles of a triangle are equal to two right angles.
The absence of the "trouble" definition suggests that perhaps it is derived from the "question" definition, and, surprisingly, very recently.
The best problems to work on are not necessarily the hardest ones, nor the ones that need the most intelligence. They're the problems you, or other people, actually have. Are you going to give up on painting your deck because it's too easy and you don't need a 500x genius to do it?
Yeah no, if you elicit Opus 5.5 , anyone else can, and you have no moat either.
But if on the other hand, I mostly use my human intelligence and just need a dumb model to complement my human intelligence at low cost and high speed (say review every commit to catch obvious bugs), I have a much better chance of building an actual moat than you do.
But outside of coding, it’s even more clear that you don’t need frontier intelligence. My customer service agent is very happy with a 100B param Deepseek flash model, thank you!
I’m using subscription models for exactly that, better models catch more subtle bugs, and they catch them faster. It works out far better in terms of work-hours saved.
Also A/ then OAI slashed token pricing by 2x~5x on their latest models
Yes, obviously a microkernel operating system that you can vibecode in a morning is not a salable product. But it's still a perfectly good operating system, and sometimes people need those, and they're a huge pain in the ass to write, especially to debug. But sometimes off-the-shelf OSes aren't good enough. Something like 50% of the embedded operating system market is still "Other/Custom".
I use Opus 5.5 daily for my job. I am aware (and in awe of) it's capabilities.
Look at the context in which I used that term 'good enough'.
What i was saying is that there are tasks for which a dumber model can be good enough, and for organizations with sovereignty/ privacy concerns, those concerns can be strong enough to incentivize the use of a dumber model.
“Good enough” as in we don’t see any point in 1-shotting everything we want to build in lightning speed. If Opus 5.5 can 1 shot it then that product is essentially commodified, no point in any one building it except as an internal tool.
If Opus can’t 1-shot it, then it must rely on our human intelligence which can be complemented well enough with a dumb model as a frontier model.
The thing is, for how long? Are they going to keep giving you "so much intelligence" for a "small" subscription dollar amount? When they really, for real, need to start making money to cover their costs, what do you think it's going to happen? Suddenly you will start having tasks that a "good enough" model is going to be fine.
I agree; yes, I can see that they need a bit less hand holding each cycle, but I also see these "frontier" agents do some absolutely dumb shit that I have to correct and then I'm wondering if I'm the looney one here.
Maybe it's because people stopped watching what their agents are doing and stopped looking at the quality of the output. But I still see agents being absolutely mindless like a junior dev.
Recent example: it updated an an API to add newly released models to the backend. There's a list of models that require specific configuration for the reasoning effort and temperature or the API call fails. GPT 6.1 Sol misses this and code fails at runtime because the newer models need to be added to the list for special handling of temp and reasoning. Fixes it for one model and tests it for that model using an E2E test. But doesn't test the other models that were added for the same error condition...I had to explicitly ask it to do so and it finds them and adds them to the list and says "that's on me."
You're not looney at all. Frontier models do dumb things all the time, especially on mature codebases. Just yesterday Opus 5.5 butchered the OOP model in a codebase I work on - it duplicated a load of classes that should have just been subclasses. A junior checked it in very satisfied that it was perfect. The LLM review passed it, the tests were fine, and it implemented the feature successfully. It's just the code design had poor taste and poor long-term maintainability.
I keep seeing this kind of thing over and over, and honestly it's not got _that_ much better since the big breakthroughs about a year ago.
For sure I happily vibecode stuff without worrying about it when it's a greenfield project, and if the LLM has written it entirely from scratch then usually it's well structured and sane. But making changes in messy, mostly human-written mature codebases is still a minefield.
Try a bigger code base or more complex stuff and you will easily see that the solution, speed and amount of problems Opus5.5 solves vs older models is relevant.
I am frequently running agents on a multi-microservice application workspace where I really need the 1M context windows, because they are filled to the brim when implementing features that require changes on several services and APIs.
This works fine with Opus 5.5. But it also works fine with GPT 6.1 Sol, Kimi K3 and MiMo 2.6 Pro.
It doesn't work equally well with Sonnet 5.5, interestingly.
I agree that Opus and GPT are almsot surely better, but so many real users are nervous enough about giving Anthropic and OpenAI access to all of their internal information that they may be willing to stomach worse models if it gives them more security.
The real question is if this model is good enough that it can still accelerate work, and not be a hindrance to real work like older Mistral models often were.
The Navier-Stokes fiasco made me push for local/controlled models very hard. "Can't rule out" that they stole data (backed up by their backdoor offers of sharing credit).
If these companies will steal from deep pockets like Disney or Sony (some of the most infamously litigious copyright trolls to ever exist), they won't think twice of stealing every bit of code you upload to them.
If your code passes through an AI company's servers, you can assume you just gave it to them. In turn, when your competitor tries to copy that new feature you just added, the AI is now trained in exactly how to copy you and eliminate your competitive edge. Unlike your employees, the AI isn't bound by the same rules and even if it were and violated them, your company probably doesn't have enough money to prove it in court (and that's if we somehow reverse some of the stupid "AI is the most transformative use of copyright I've ever seen" judges who have drunk the coolaid).
Most companies could build the compute to run GLM or Kimi models for way less than the potential loss due to IP theft from using third-party systems.
I would also add to this, there are ways to use customer data to improve your model outside of just using it as “training-data”.
A simple loophole, use the code to create an RLVR environment where the resultant code is the end goal / max reward. Technically the customer data is never trained upon, but effectively you’re using it. Even better, use the code as a seed to generate synthetic data similar to it and use that synthetic data as rewards in an RLVR model.
Unless you can host the ChatGPT model on your own servers, which I know some enterprises are doing, I don’t think there’s any hope of protecting your data / competitive advantage from these frontier companies. Better to be paranoid, than be commodified by these companies.
tbh to me if the AI company writes all of your code & your eng don't even review it anymore then… the AI company _controls your company_. maybe that's ok if you make widgets but less ok if you do anything in dev tooling, security, or [insert market they may suddenly decide to compete in].
> Following an investigation, we have confirmed that Buckmaster’s Codex prompts over the two months preceding this announcement and paper on September 8, 2026, could not have influenced the system in any way, including through training. The OpenAI internal model used for this result was developed through large-scale reinforcement learning on top of a previously pretrained model. Our proofs also differ significantly. In the Euler case, Alpöge and Buckmaster proved a result with external forcing, while OpenAI’s system proved a result without external forcing.
It's an improvement, but game changer might be a bit of a stretch. If I lost access to Anthropic or OpenAI models tomorrow, I would be annoyed, but would reach for a slightly inferior model. Last year I wouldnt be able to say the same, and rhe challenge is that the moat is drying up fast. Whether its general improvements in model training by other competitors, or straight up distillation of SOTA models, the moat is shrinking and the available capital and spend for American model providers is going to dry up quickly as competing good enough models are adopted by more consumers.
It's especially the case as more non-Americans look to self hosted models and domestic cloud inference providers using open models that the US providers who are still leading the charge need to drastically drop their prices and find a path to profitability in order to maintain their lead and retain the advantage they had as AI turns into a commodity (which is happening faster than I think even the frontier labs initially predicted).
Please, give it another 6 months and they catch up. The American labs are currently trying everything they can to block others instead of advancing their models, trying to build an artificial moat. The American models are not that great, they are good, and they have a lot of agentic workflows in the back, but its basically a hardware limitation at this point. Once the HW makers catch up, and we can move away from the Nvidia monopoly, things will speed up quite a lot IMO.
We've been hearing the line about them only being a few months behind for a year now, during which time O/A have grown their revenue like 10x, haven't they?
We've also been hearing we're 6 months from AGI for about three years, and here we are.
"Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!"
Nope, wrong bit of slightly-surreal children's literature. Through the Looking-Glass and what Alice Found There by C L Dodgson, otherwise known as Lewis Carroll.
Those are two different things. The market is expanding, so even if competitors are catching up, you can have your own revenue, in absolute terms, grow.
That's fair, I left out a part of what I keep hearing in relation to the open weight models only being a few months behind, which is that they're going to eat the big labs' lunch, but then the big labs keep growing...
Yeah, I could buy a bigger and bigger vehicle when a new one goes to market, but using a semi-truck to drive myself around would not only be significantly more wasteful but also more inconvenient and limiting compared to a sedan. Very much like advanced closed models vs performant open ones.
A lot of big enterprise customers are already starting to question their spend and there is more and more pushback in regards to sharing private data with both AI orgs, even with the enterprise agreements in place. There might be growing revenue at the moment, but enterprise finance moves slowly, so the next couple of years will be the deciding factor.
I do a broad amount of diverse experiments/projects I always wanted to do and throwing Opus5.5 against it just works
I have to admit, Sonnet got really good too.
But Opus just uses tools, a broad spectrum of it, etc. it feels like sure if you add some router behind it you could split it up if you need to but if you give me the choice, its opus allll day long.
I hear this literally every other week about whatever the newest FoTM model is.
Unless you can provide concrete examples of things you can do with them that you simply couldn't do with last week's model, it's absolutely meaningless.
Yeah, I agree with this. I think the "the models are good enough" narrative is a myth. I've heard it so many times over the last year, but the model number keeps changing...
There is no ceiling on what you can accomplish with more intelligence, so there will always be a market for the best models, and that market is likely to just keep growing. If Opus 13.5 can one-shot a profitable company or discover a new disease treatment or whatever you can think of that a swarm of relentless super-geniuses could accomplish, companies (and governments) will throw money at it.
I also think there will always be a market for many sub-frontier models that will continue to grow rapidly as well, because "good enough" is definitely a thing for a given task.
One could still argue that models are good enough for a given task. I primarily use Opus at work for writing code and I realized that for my usage the intelligence of Opus 4.8 is more than enough. Sure the newer models are better but I can still do my work with having access to newer models
It will become winner take all when AI companies manage to really get value from user logs.
Right now they don't even get good feedback from local sessions - I can see it make the same mistake two days running, and then months later when a new model comes out, presumably trained on my data, it still makes the same mistake.
My theory is there is actual knowledge in the user logs.
When a user says "I'm struggling to undo a bolt on my 1952 Mustang" and the AI responds "try hitting it with a hammer" and the user replies with "that worked thanks" - that is a tiny piece of knowledge which exists nowhere else.
Future AI's can say with more confidence that hitting it with a hammer will probably work.
Across billions of conversations, that can add up to more knowledge than all books.
Truly, this is what the Lord's prophets have revealed to us! (Eliezer 11:52) Keep strong in your P(singularity), for when the Kingdom arrives, He shall judge us in His righteous glory, whether to eternal annihilation, or rebirth and life in His Memory Eternal!
Assuming RSI is something that is possible as you envision it in the near term. I think that it will happen at some point, but I think we could still be a long way off. I don't think anyone can truthfully say that it is right around the corner.
Especially with Mistral taking a fraction of the investment of the big guys. They can maintain the position pretty comfortably just by staying within a standard deviation of the leaders.
> Its quite interesting to see that at least the early days of AI so far have not been a winner-take-all runaway acceleration game where catchup is impossible.
Mistral is also an European company. As we live in a time where the US regime is engaged in pyrrhic geopolitical tactics, it's good to know that it can't threaten to cut access to models during s period where everyone is rushing to incorporate them more and more in our life.
On a purely technical level, maybe? But in terms of actual revenue, is there really any chance of anyone catching the big labs?
Obviously, this is only a valid question if you don't believe that open weights are about to eat their lunch and their revenue is about to collapse, or they're running a super unprofitable ponzi scheme propped up by investor money that's about to collapse like a house of cards. I don't find those positions credible at all though.
If you do, then this question isn't really for you, as I'm more interested in thoughts from those who think that OpenAI and Anthropic in particular are about to be the largest companies on earth in a couple years. Could anyone catch them at that point?
> But in terms of actual revenue, is there really any chance of anyone catching the big labs?
I don't know about revenue, but I suspect multiple other labs are already beating OpenAI/Anthropic on profitability. Staying on the frontier is expensive, and it's hard to recoup those R&D costs when you have a bunch of other labs nipping at your heels.
If you concede the previous point, then the only way for OpenAI/Anthropic to keep growing long term is to swallow the whole economy (i.e. mass job replacemnt), and that's a bet I wouldn't take.
I think the actual plan is to swallow a good portion of the job market. It’s the only thing that makes sense and I hear VC podcast debates on which percentage of jobs justifies the market cap.
Maybe. I can't freaking wait for the IPO filings so we can finally put all this to rest. (haha, like that'll actually put it to rest on HN, but at least we'll have better data)
They are very unprofitable…? I don’t think that’s really in dispute. We haven’t yet seen a profitable frontier lab and model pricing remains fairly subsidized
Seems silly not to have predicted that 10 years ago. I feel like it's long been obvious that smarter models being available will mean way easier cheap synthetic data and access to tools that will speed up competitors as well as consumers.
Man, a lot of this discussion sounds like people cheering for the last kid crossing the finish line.
Surely we want competition and Europe involved in that, but at this point I have grown used to either American labs smashing the frontier remarkably fast, or Chinese labs getting way, way closer than you would expect them to.
Mistral’s progress, regrettably, feels much slower. This model doesn’t knock anybody’s socks off. The model is (and I hate to be this harsh) mediocre, and this mediocrity has also arrived months late.
> Man, a lot of this discussion sounds like people cheering for the last kid crossing the finish line.
Sometimes it's ok to cheer for the last kid crossing the finish line because they're actually running a totally different race, and winning might look completely different.
When I look at what Mistral does vs other organizations I'm impressed:
They aren't profitable yet, but they're a lot closer than most and they're doing a hell of a lot with very little.
Pointless racing story:
I was in high school track with a really tough guy who was just not a runner. We went to a pretty messed up high school and if you screwed around in track practice sometimes the coach would make you run a crap race at the next meet, like steeplechase or hurdles. Well this guy and a few others screwed up and coach made them all run hurdles at a meet.
He hooked every single one and fell on his face. Every time he got back up and kept on running. By the time he hit the finish line his knees were bleeding halfway down to his ankles. We cheered like hell and he was smiling ear to ear.
Mistral is more profitable as a general percentage sure, but take their overall revenue: about $400 million annually.
Good, until you compare that to OpenAI’s behemoth $70 billion.
Sure OpenAI has much higher operating expenses, but which position would you rather be in? If OpenAI cut off all training and R&D tomorrow, would their 1 billion users simply vanish? Could Mistral ever realistically acquire 1 billion users?
Point being that profitable isn’t telling the whole story. And I don’t know if lack of resources is a great excuse, when Chinese labs are doing more with less, and it’s ultimately up to every competitor to win as best they can, regardless of their circumstances.
Great story you told, though. My old track coach was also an asshole, but of the more lovable variety.
I think right now it should definitely be a race to profitability, because revenue will not save you from unsustainable finances, no matter how large the values are.
I think OpenAI is currently far enough ahead that they could just cut training and R&D for a while (maybe even 1-2 years) and rake in some money before they'd need to start again, but if a smaller company finds a way to keep doing R&D while also being profitable, it doesn't matter if they're a little behind in quality, they will win in the long term.
I read that site quite differently from you. You seem to be analyzing absolute differences but ROI is really about ratio of spending to revenue.
It looks like Mistral is middle of the pack, behind Anthropic and ahead of OpenAI on that front. All of those labs are way "ahead" of the cloud providers, but those providers are building infrastructure, not just training models, so it's not apples to apples.
Proportionally to income, Anthropic revenue is two thirds of its expenses, Mistral is a half, and OpenAI a little bit than a half. They aren't that different.
This is a great story. I don't think we should expect someone to win in the first races they participate. There are football teams that never win the championship and they always play and strive to be better.
Mistral it's definitely improving with each release and props to them for releasing open models
> This is a pretty grim prognosis for European AI.
I think it's an incomplete read. What's the point in competing for a sizeable percentage of your funding when the finish line is incrementally being moved each month? Better spend it on leapfrogs which they seem to have done.
Meanwhile Mistral have a natural ace in their pocket with respect to regulation in the form of CADA and the Cloud Sovereignty Framework. I can't think of another company that would qualify as SOV-3 under that regime
As opposed to what? Companies that steal the collective intellect and intellectual property of the entire world? Yeah, what a great alternative. A world where all of your hard work is blatantly stolen.
Ah yes, a red herring, plus my favorite HN hypocrisy, all in one comment.
The 'information wants to be free' crowd complaining about protecting intellectual property now that it threatens their income stream...has to give the patent trolls and the RIAA a delicious bit of schadenfreude.
I invented this react button style with 8px border radius, ChatGPT stole it from me, I want rents!
People were extremely dismissive of chinese models until recently. They went from 1 year behind frontier to 6 months behind frontier to 3 months behind frontier extremely fast.
Important to point out that these 'X months behind frontier' really refer to the public frontier, and not the actual frontier, which private companies are free to protect indefinitely. Perhaps open models are in actuality 18 months behind the actual frontier - how would any of us know?
Because we have access to these models and use them maybe? Because of benchmarks? It does not feel like there is much moat, look you can even use small models these days and do productive work. It is rather that private ones starting to get into diminishing return area which without moat is quite risky.
Hey I’m cheering too, but this kid is less like a kid and more like a multiple gold medal winning Olympian, and our once great track star is now on his last legs.
Europe has been at the heart of technological innovation for a long time. In recent years, this feels much less so.
This is a pretty mean spirited comment that fails to recognise the realities on the ground. The US is a technological juggernaut which Europe just can't compete with, because it's not a level playing field. The US has multiple moats in the form of financing, computing power, tech talent and one language.
This is very much a welcome development and we should celebrate it.
I don't think it will matter in a year or so. We are clearly topping out on useful intelligence for an increasing amount of tasks, as demonstrated by more and more models reaching the "useful" barrier.
This barrier is not going to start moving dramatically. It will simply be mostly satisfied for most work we do. Mistral is going to get there, soonish, long before the economy takes an entirely different shape (in so far that even happens).
There will be super human intelligence tasks, tasks truly constrained by intelligence for quite a while. Those will be few and far between, relatively speaking. Mistral will have plenty of opportunity to capture the other stuff, with a fraction of the resources required that it took the frontier labs to get there first.
It’s not that I necessarily think you’re wrong—-it actually makes perfect sense to me that eventually utility from frontier models will have diminishing returns—-but this has been a talking point for a long time.
The prediction that eventually more intelligence will stop being significantly more useful has been wrong many times before at this point. We should be prepared for that not to be the case.
I think it's reasonable to expect much diminishing returns for work that we have a good sense of. Compare it to a new laptop, phone or machine in a factory: There probably are better versions than the one you are using, but how much better are they and will it help you do better work? Is it even enough to be worth switching?
(I understand that the considerations are a bit different when ostensibly the switch is as simple as changing an API call, but that's not really true, when you have the AI integrated into a larger system and need to run evals and adjust the framework around it at some places at the very least)
Then, of course, there will be entirely new work that benefits from every bit of additional intelligence, but the stuff that is useful to 99% of the economy right now is probably capped on some relatively finite amount of that (with adaption/integration going to be the bulk of work)
Like I said, I don't disagree with the reason of your argument. Rather, I doubt because of a certain epistemological humility.
We can sit here and predict things all we want, but these sort of predictions don't have a great track record. The way things actually play out with AI advancement is often different from the way our perspectives reason they will play out.
It's better then to rely on previous experience and evidence, rather than using reason alone.
Curious, why do you say the model is mediocre? I haven't tried it, so I can't pass any judgement... I've learned to distrust benchmark rankings. Are benchmarks and Artificial Analysis the yardstick you're using?
> Curious, why do you say the model is mediocre? I haven't tried it, so I can't pass any judgement.
It's nowhere mediocre.
It's toes-to-toes with GLM-5.3 which is one of the best Open Weight model available (With Kimi K3) for general reasoning.
I just runned it on code reviews right now and it was able to catch some thread safety issue than DeepSeek-4.1 didn't. And DeepSeek-4.1 is by no means a bad model.
I'm just wondering if it's really the result of impressive innovation, in line with comments saying that Mistral might in the future beat a bankrupt OpenAI due to being more profitable or efficient.
Or if it's more a situation where I could have Opus 5.5 go through recent papers from the Chinese labs to apply these to a new large training run, resulting in a model that performs similarly to Mistral Large 4.
Any lab can make the lack of resources argument. Regardless of that though, it’s every lab’s own responsibility to do well, and to do well despite circumstances.
Is that fair? Of course not! But who really expected for everything to be perfectly fair all the time anyways?
We can do more nuanced comparisons that take resources into account. That will help us see how efficient each lab is, but it doesn’t help the labs overall. Mistral shouldn’t be held to a lower standard simply because they got outcompeted.
Sure sure but if I want to answer the question "is model quality a function of resources or there's something irreplaceable in the US and Chinese engineers that can't be replicated in Europe?" it's a good metric. And if the European Union or an European VC is going to decide whether to invest in AI in Europe or not, they'd probably like to know.
Doesn't matter, it's excellent, and it's European with European inference, which solves the pains of all my clients trying to build data lakes and processes on top of it.
Nobody in the real world cares about minor benchmark differences in money losing coding agents.
And nobody in the real world is giving Altman or Musk their data.
I wish you were true - but I still meet a lot of people handling sensitive data and using free or cheap version of ChatGPT or Claude with their customer data.
I think we will see some horror stories come out with data leak in the next years.
Yes, it arrived months late. The waiting was too long I started losing faith. But when I saw Mistral’s series funding news, I knew they are on the right track. 3 billion EUR is just too significant
Those are comments from Europe. The US is waking up now and I expect them to be much harsher.
I really want them to win as that's our last horse in the AI race, but ~200 research-oriented devs out of 1800 employees? I believe they agree it's pretty doomed and have pivoted.
First few models will always be slow improving and worse. The way to improvement is working your way through a gajillion evals [1], finding bugs, gaps, and curating training data (this part involves human design as well as raw inference compute) to fix it. This is very time intensive and can't easily be "done once and then everyone has lesser work to do" since every model is different. Well, one way to accelerate it is to simply have more compute, which mostly openai and anthropic have[2].
This is mistrals first 1T-scale model and I expect the 4th or 5th generation to be close to the best for many purposes.
[1] These evals differ from the public ones like terminal-bench, are sometimes model-specific, need real, diverse usage to actually create, and are held secretly since quality of eval is the first driver behind the next step improvement of a model.
[2] It is not close. This model was trained on less than 4k GPUs, whereas astra used north of 100k GPUs.
Mistral is not that new a player though. How can we give them this much grace when other players like xAI have done more in even less time? I don't think coddling Mistral helps them.
And to the point of scale and training cluster, so what? Not only do Chinese labs have smaller clusters with less empowered GPUs, compute is Mistral's responsibility. You can't take away from other labs just because they fulfill that responsibility better.
xAI has a lot of compute. Deepseek also has a lot, not as much though. But this is changing with their new 160k huawei ascend datacenter in inner mongolia.
The lack of compute is not really attributable in that sense to mistral. First of all it needs general investor and government willingness, which is easier in a larger economy like the US or China.
Second, you need widespread usage of your paid inference service for two reasons: one it pays off your compute cost, and two it speeds up the improvement process.
The vast majority of deepseeks paid customers are within china itself (since openai and anthropic services are not reachable from china) which gives it a market. But for someone in france, there is no reason to use a structurally slower developing model from mistral compared to using one from openai...except when data guarantees are needed, hence the landing page focus on sovereignty. As far as the dual use aspect goes, a model like this is more than enough, so the government will be happy.
My point is that pretty much anybody can make the case that getting compute is hard. Regardless of that though, it’s every lab’s own responsibility to do well, and to do well despite circumstances.
Is that fair? Of course not! But it’s not useful to remark on parity, since nobody ever expected everything to be perfectly fair anyways.
We can make all the excuses we want. All it results in is holding Mistral to a lower standard. Does that result in more fair comparisons in regard to resources to output ratio? Maybe. But in my opinion, this sentiment is detrimental to European AI overall. We shouldn’t simply accept European labs as perennial losers, or normalize lower expectations. That doesn’t help anyone.
It's a bit below DeepSeek 4.1 Flash at about twice the size. For a model that was supposed to come out a few months ago this is pretty good. Mistral catching up to the chinese open-weight models is great news. Excited to see how they will build on that!
What I'm looking for with this is of it hallucinates less that DeepSeek with its sparse attention, and can be used in better summarization and content generation in multiple languages.
DeepSeek sucks with languages other than English and China.
I noticed they compared with the previous best deepseek model, 4 pro 0813 for the two coding benchmarks (DeepSWE and Terminal Bench) and not Deepseek 4.1 flash. A glaring omission really.
Very nice progress. Also I respect putting Kimi on those charts. Regardless of if they are beating the Pareto frontier (not now), model diversity is a good thing for humanity — I’m hopeful for the team to keep increasing their gains.
I'm literally English and I feel the need to defend the French here, they're among the most successful military powers in the world historically speaking. It's hardly fair to judge a thousand years of French military history by the outcome of one war where they didn't do well.
As far as your sense of humor goes, perhaps you can find a different niche than disrespecting the memory of the 100,000 French soldiers who died in the early weeks of WW2 precisely because none would surrender? That 90s revisionist trope to shame the French for not being neocon enough is beyond stale.
I’d rather have one European AI lab like Mistral, with the financial firepower and compute to pretrain its own large models, than 20 weak labs that fine-tune Chinese models.
AI needs big bucks, and Mistral is Europe’s Anthropic.
Absolutely not, the last thing the Europeans need is a monopoly that would turn into a bureaucratic moat. Competition is always good and new startups should get the chance to challenge the ones that cannot compete.
it's extremely unlikely that they re-use anything from a chinese model, that would be obvious quickly, what's more likely is using documents produced by a better model to create synthetic data.
The US has made sure that Mistral has a large market in the EU by temporarily preventing non citizens from accessing Fable.
A lot of European companies now want a model the US can't cut off, but also lack trust in Chinese models.
Some of these will self host Mistral but most will pay them by the token. It's not going to be a huge market or a huge margin within that market but probably it'll be enough.
This is the exact mentality that makes the EU fall behind. If you don't want to invest in something until it makes a profit, you don't get the benefits of being a pioneer.
> Europe needs profitable AI companies, not money pits.
AI is a strategic technology with obvious national security implications. EU should invest in its development whether it's currently profitable or not.
there are no profitable AI companies at the moment ... This is the exact problem of European startups, trying to make them profitable from day 1 while American counterparts (and Chinese) keep bleeding money for years. Europe will never have a Tesla, a Google or an Amazon with that mindset.
Google treats its customers the same way as Amazon treats its employees. Any sane person stays away from its services because a bot can close an account for no reason, and there will be no recourse. A total shit company, like some illegal IPTV provider: no accountability, and when the service disappears, you have nobody to turn to.
As if European companies are any better. Have you tried getting customer service from any European ISP for example? And companies still treat their workers like shit, it's just you can put up more easily with it since you can't get fired immediately.
Ah yes. Let’s thank the might US for providing pesky Europeans capital a fistful of dollars. But maybe let’s do it after Americans thank for Russian, Arabian, Chinese, European capital and workforce. After all this is what “Made in the USA” means.
That said I love that they don't seem to restrict cyber capabilities to any degree and even lean into it.
If the best model for cyber attacks is open for everyone to use it just makes us all safer I think. Of course you then also HAVE to use it or otherwise you're vulnerable, which is a great distribution play.
Too late to edit, now including AA-omni Score [0] and 'by Domain' Software Engineering [1]. Also added comparison to the eyeballs-median of the 'top 10' models and then you see that indeed Mistral Large 4 Preview scores miserable in the AA-Omni indices.
Inte Cost AA- Omni
llig per Omni Softw
Open Weight model ence Speed Task score Eng
Mimo-V.26-Pro 46 47 $0.13 8 33
GLM-5.3 (max) 45 73 $2.01 14 37
DeepSeek 4.1 Flash Max 39 227 $0.27 -5 34
Mistral Large 4 Preview 38 116 $1.13 -5 5
Closed/proprietary ~50 110- $1.50- ~43 ~85
median top 10 242 $7.50
I ran it on my trivia game Redactle which features a redacted Wiki article. Mistral Large 4 is not very good. It can sometimes solve a game with ~40 guesses whereas the top models like Gemini 3.8 Flash or Grok 4.7 can one shot most puzzles. My benchmark here aligns with AA Omniscience. I also have a version where the text is rewritten to detect over fitting to exact wiki text which changes the scores but not the leaderboard order.
Gemini models are summarizing wikipedia articles all day (when being used for Google's ai answer), can we draw some conclusion from this, did they train it extra well on wikipedia content?
>If the best model for cyber attacks is open for everyone to use it just makes us all safer
Issue is..
I don't believe for an instant that any of us, including US citizens, get access to the best models for cyber that the US has. I think any adversary would have to assume the models in use by the US side are unreleased.
US is not the only one dealing under the table by the way, I also think everyone should take China having unreleased models as an operating assumption at this point.
So Mistral is the best that the public gets access to. And that's if it's even the best? Benchmarks and pragmatic work have often been shown to be two radically different things in this industry.
It's a joke, chonky is used to refer to fat cats, and there was a joke meme over the summer that Mistral are going to release a new model, Le Chaton Fat (chaton is kitten in French). The name is a nod to the memes.
Only model Pontifex considers a possibly conscious being. Obsessed with jumping in boxes with Cesium and some poison though, even escaped the litter box during testing, probably an alignment issue…
I've read it a few times and I always processed it as "Le Chateau Fat" ("The fat castle"), which I guess also kinda works. I guess it is mostly because my french is so rusty I forgot (or maybe never learned?) what chaton meant.
The main thing I always get away from the comparison tables of these "big" models, is how well Deepseek v4.1 Flash performs. While still being the cheapest model by a long shot.
Beyond benchmarks, does it in day-to-day? Have always struggled to get competitive performance out of any Deepseek release going back to V3 vs Z.AI and Moonshot models. Maybe I really suck at whatever is needed to make DS models fly, but even tailoring my suite hasn’t gotten me far when I tried with V4 Pro. Happy for anyone who is able to leverage their models well, wish I’d be able to crack how to leverage them.
Will say their research is some of the best reads in the industry and I could not care less about their model release cadence as long as papers keep coming.
Yes. V4.1 flash performs really well. I don't know what to say, maybe you don't believe, but my team has been using it mainly for almost a month now for programming. It is as bad and annoying as any of the US sota, but costs pennies. And if you look close enough you find providers that can push it 300-500 tokens per second...
Maybe it is due to us being all very experienced devs. And can steer the model. But my daily routine is just to have 8-9 Zellij tabs open, DeepSeek in omp in each, and grind research and code day and night. Really nice model...
Hold up, 300-500tps at p50? At p99, even something like Opus 5.5 on fast reliably hits 330tps+, so that'd be expected but if truly p50, wow. And no regressions in tool call and structured output vs the official endpoint? Would love to test that with my evals, please share.
Yeah, very similar benchmarks at 1/4 the price. I've been pretty happy with Luna 6 though I still think the gap between small and frontier models is larger than many people want to admit.
It is not super good in long horizon tasks and worse than 5.6 in our evals. It really failed the agentic evals where DeepSeek, Kimi and Opus are the winners.
I'm at a loss as to what to do now. I've been wanting to support Mistral for so long. I struggled on with Mistral Medium 3.5 for longer than I should have (although I also think it taught me some valuable process lessons).
Recently I switched to the Mistral hosted GLM-5.3, this worked very well and powered through a tonne of work. Unfortunately, I also completely maxed out two subscriptions within the space of six days this month. One can't stack subscriptions with Mistral, so I'd have to register a third account for another subscription, which will be annoying with changing API keys all the time. Sure I can switch to pay-as-you-go API, but that adds up really fast. The Mistral dashboard shows that a Vibe CLI monthly subscription for €18.44 actually provides €255 worth of API use (apparently, and I tried to check this with Support but it seems like they were intentionally vague).
After maxing out my Mistral subs this morning, I dropped $10 on Xiaomi to try MiMo-2.6. So far so good, seem to have done a lot of work for the $2.85 I've spent, and Xiaomi prices are still much better than the Mistral introductory offer for Le Chonk.
Not sure where to jump.
Edit: Not being able to stack subs is my biggest gripe with Mistral. I'd probably pay them $100 per month (5 subs worth), but I'm not going to switch to the pay-as-you-go API and burn much more money for the same amount of tokens. Instead, I've taken that extra money elsewhere. If they just allowed one to keep topping up subscriptions on the same account it'd be grand. Or even a bigger single subscription. Make a $100 tier with five times the capacity.
> Recently I switched to the Mistral hosted GLM-5.3
> [...] I also completely maxed out two subscriptions [...]
TIL that Mistral provides hosted GLM-5.3 and that it has a subscription model as well.
> Not being able to stack subs is my biggest gripe with Mistral.
If you were to use a gateway like GoModel[1] then you could use Virtual Models[2] to load balance across the subscriptions. For that to work, though, you would have to register the same provider as many times as subscriptions you have.
It can all be defined in code, so you'd only have to do this once.
> Not being able to stack subs is my biggest gripe with Mistral.
Haven't tried it myself, but on Discord they've recently discussed this: Sign up for a team plan with multiple seats, only use one seat, usage limits are shared. Not against TOS.
Yes, but I still don't know where the truth is.... The Mistral dashboard shows a drawn down on €255 worth of "credit" when you have a Pro subscription, which costs €18.44. There's no way I'm going to use the pay-as-you-go API if it means I'd be burning €500 in the next six days, as I just ostensibly have in the last 6.
I appreciate I can set a monthly spending limit for the pay-as-you-go API, but I'm just not willing to find out how far €100 will go, when I know it will go further elsewhere.
I just wish they had that €100 subscription tier, for the equivalent of €1k pay-as-you-go use.
I am on the same boat. 3.5 medium was disappointing, glm 5.2 and 5.3 were very welcome. But I still cannot leave anthropic or openai. This chonky model being expensive, mistral’s subscription quota will run out ridiculously fast
Seems hard to get customers at that price range when you're competing with open source models that are 1/2 - 1/3 the price but with similar capabilities.
People usually buy the cheapest, like Deepseek or GLM or they spend on Anthropic/OpenAI subs.. Are these models in the middle getting any users?
On a side note, I wonder if this was the popular free Space Bunny model that left openrouter yesterday.
If you are a EU company worried about your data then mistral is your only option. Think about EU military companies. They can't use US and Chinese models.
If you want to use the model for something, where hiring a Chinese national to work on these tasks is a no go. Using a Chinese model is going to be a problem as well.
Some tasks and industries are so sensitive, that countries will not allow you to risk, that the model is aligned with Chinese interests and not yours.
I think you misunderstand how data processing works in a LLM. You absolutely can download the weights of a chinese model and run it on hardware you control.
I've been dreaming of this for a simple reason: the french prose combined with GLM 5.3 reasoning capabilities.
GLM 5.3 is incredible because for the first time with an open-source model, it feels.. enough. I don't need much anymore, this model is great in everything. Except a thing : speaking french.
If the benchmarks are true, I'd be glad to switch entirely to Mistral.
You are quoting the discount pricing. It is 50% off for the next two weeks. The blog post has the real pricing up front in the card on the right side: https://mistral.ai/news/mistral-large-4/
After that, it will be much closer to GLM 5.3, but you can also get 5.3 in their API! I dont see people really talking about that.
Excited to try this. The low costs v. benchmarks alone here are worth a serious test. K3 has been my daily driver for a month or two now and it's dramatically reduced token spend (while not having much of a negative impact on productivity).
This was the era of the AI race I was waiting for.
Haven't tried it yet. It looks like sol is a hair more expensive on input, a hair less expensive on output ($2/in, $10/out per 1m, K3 = $0.82/in, $13/out per 1m).
Nice! Once they make it available through their API I will be happy to support them. My local Qwen3.8 27B is serving me well, but I miss the speed and concurrency that comes with subscriptions, and I am not currently paying for any.
Curious that it scores higher than opus 5.5 in cybersecurity because the closed models refuse to comply. I wonder if that means it's more susceptible to offensive uses.
There was a section showing malware analysis which is great to have. I suspect everyone is using Chinese for this nowadays but now you can switch to this as well. Opus like GPT are restricted, it is hard to work with them doing offensive security and Fable is expensive.
All things considered I'm more inclained to pay for EU-based AI in the end (supporting local company and most likely being more aligned with EU regulations…)
I think it is important what they mentioned about using languages as well. Some smaller nations think they will culturally vanish because their “heritage” is close to nonexistent in the training data. If Mistral uses more of that language specific data it is a good thing and might help selling it to different non-english groups.
Maybe I'm missing something. Doesn't seem super impressive to me. A proprietary model with performance comparable to GPT-6 Luna and Deepseek 4.1 Flash, but at a higher price than either. The main selling point is that it's made in Europe... not very compelling, globally. I suppose maybe there is some niche where European-hosted open-weight models aren't enough to satisfy some EU regulation, where only the use of European-trained models is in compliance, but as a non-European I have no idea what that niche would be.
Side note: Wish this thread was more focused on talking about the model instead of debating about China and America. Whatever happened to staying on-topic?
You yourself also kind of pointed out why the discussion about USA and China is not off-topic. The niche Mistral wants to fill (afaik) is that it's European. I, as a European am really happy that they made such progress in so much worse (financial) conditions. I guess it's mostly geopolitics.
I take that back, actually: I didn't realize they release them under a modified MIT license requiring large corporate users to enter into commercial agreements. So it is in fact proprietary. Sorry for spreading misinformation.
I created a brand new programming language to test LLM models on[2]. It looks like Mistral Large 4 fares about as well as Kimi K3 and Claude Sonnet 5.5[1]. Though it is much slower than Sonnet 5.5 and more expensive than both of those.
Well, with this and Beam people are going to have to stop saying that western open models are dead. This is great news. Anthropic and OpenAI may have a bit more knowledge, talent and compute but they don't have a monopoly.
Investors looking for them to make monopoly profits are going to be disappointed The premium they can extract from consumers for their models will be capped. Tokens are likely to remain close to the cost of compute, a cost which is high but falling fast.
Also, yay Europe! Although the comparison between this and mimo 2.6 is not flattering...
Sans context, I really like the Anthropic and Claude's faux-academic, minimal-ish, intellectual-ish branding.
(Especially the way it looked ~12 months ago -- it's gotten more cluttered since then. Perhaps unavoidably, as the breadth of their offerings has grown)
But over time it's begun to feel like unsettling cognitive dissonance as their ambitions grow and the stuff to worry about has piled up.
I know the whole “AI logos look like buttholes” thing is a joke for most people, but I really speaks to the pornification of our society. I would never have thought “butthole” looking at any of their logos.
I think there's such a thing as throwing too many marketing and sales people and too much polishing and "refinement" at something. I see in new product announcements from Microsoft as well. It's like seeing someone try too hard to impress you.
I do my best to avoid any AI marketing because I extremely despise it.
I just haven't quite decided on my new profession, yet, but it's either going to be with plants or with animals.
Give them time. ML 4.0 was just pretrained. Mistral will certainly use it as the base for distillation and RL for smaller, better, more efficient iterations, just as the competition does.
Pretty interesting but I wonder apart from sovereign use-cases in EU, who uses Mistral and for what? How good are they compared to frontier open models like Kimi?
I agree sovereign markets themselves can be huge but just curious.
There is no model passing all the tests. Gemini is on top because questions also include general knowledge, trivia or domain-specific questions, which Gemini models are way above any other model.
Even if models have similar scores at the top, it's useful to compare costs/response times.
Everything in the top 25 is above a 9.4 out of 10, so it seems a bit saturated?
Also seeing that Claude Opus 5.5 is at #23 makes me think this is not the benchmark for me
It's actually quite hard to make a single task/question with few steps max that SOTA models fail on. Gemini are on top, because asking domain specific questions still breaks most top models, but not Gemini ones.
Claude is lower, and also most Antrophic models, because (historically) they have one of the highest refusal rates and also struggled more with precisely following instructions (i.e. outputting in a specific format/way).
You can assume that all models above a score of 9 are "good", then it's more a matter of preference for their response times, costs, personality and strength in specific areas.
Opus 5.5 are best at coding, and even art generation, but struggle a lot with built-in knowledge, when web search is not available.
> The RL run behind this preview is still in flight and shows no sign of saturation -- we will release a final version before the end of the month along with the weights of the model.
Does it have the ability to capture the market like OpenAi or Anthropic ? My point is, Regular/Average users of AI do not really care about benchmarking. Marketing decides which company makes it to the phones or PCs.
> Marketing decides which company makes it to the phones or PCs.
Right now a big part of LLM market is people using it for professional software development. Most of these users probably care about the quality of the model and also notice it during daily work.
For normal consumers, shure it doesn't matter. In the end the ai summary of google will probably be the most used as they are already exposed to it anyways.
It's more than a sales pitch. Given the current geopolitical climate if I were int the eu I would not trust any us tech company with anything. It's possibly one of the most important things they can do to survive the next 5 years.
Still second most expensive open-weight model. I don't care about cybersecurity index.
And still can't beat Chinese models but good to see European in the game.
I think it's actually a wink and a nod towards the social media meme of "Le Chaton Fat," a fictional model that is jokingly attributed to Mistral, usually with century-defining benchmarks and unfathomable size.
I tried it a bit and I like it! It is very fast via openrouter (significantly better than Kimi K3) on webui. Very verbose and starts to forget instructions after awhile it seems, but it gave me quite a lot of good info during a half an hour chat on C and embedded programming. I think I will keep urimg this as my main assistant for few weeks.
Excited to test this on my benchmark[1], but I'm guessing it won't fare better than MiMo v2.6 Pro which is currently the best open-weight model as far as I'm concerned.
Why isn't this on openrouter yet? Is there a better router that gets models much quicker?
I tried Mistral's last 3 large models, devstral and mistrallarge3 and the numbers were not even benchmaxxed, but just false. the models were so weak and garbage. Let's hope they are telling the truth this time around, we need more alternatives. There mistral-small and original MistralLarge models were awesome, hoping they are back!
It also has guardrails, albeit very weak (which you can read in thinking traces).
I've had seen subagents fail with "hmm, hmm - hmm, hmm (...)" multiple times, it keeps trying to cheat at the task (trying to remember hallucinations).
GLM 5.3 (which Mistral also serves at 3x the speed!) wipes the floor with it
I tried a quick abbreviated kimbench on their playground before bothering to do the whole thing.
Maybe I didn't really select mistral 4? Either way, failed completely on question 1 and the next 2 questions were completely off base too. I didn't bother to finish.
What experience have people had with Mistral models for cyber research? are they as constrainted as Anthropic models? I use claude day2day but have the need for a model with fewer guardrails to pentest our own APIs.
They explicitly lean in to cyber work, and they appear to be very permissive from their marketing:
> This is particularly important in cybersecurity, where provider-level refusals can block legitimate vulnerability research and incident response, and where losing access to a capability mid-incident can itself become a critical security risk. ML4 pairs top-tier cyber performance with open weights and self-deployment, giving organizations both the capability and the autonomy to run advanced security work under their own policies.
> That top score reflects a practical advantage. Several leading closed models, including Claude Opus 5.5 and GPT-6 Astra, score near zero on the same test because they refuse to perform the task. Yet defending software often starts with proving that a flaw is real, exactly the kind of work safety filters in closed models can block. This matters even more as threat actors increasingly jailbreak those same models to support offensive cyber activity
Is the reason for the massive gains in certain benchmarks due to distillation from the other lead models hence the slightly "under" pattern seen in the comparison charts?
I tried testing it, but reasoning effort parameter doesn't seem to be there and output is sort of broken because it reasons directly in the output tokens...
Seems like they've finally made a genuinely competitive model since the original LLM craze, congrats to them! Glad to see some diversification in open weights providers.
All things considered I'm more inclained to pay for EU-based AI in the end (supporting local company and most likely being more aligned with EU regulations…)
This sure is nice, but I've had less than satisfactory results with GLM5.3. I'd like Mistral to compete with Qwen3.8-Flash-Next a 120B class model that IMO is the first model that I can use for serious coding while running it locally.
I estimate it's coding ability on par with opus 4.6 (but opus definitely beats it on factual knowledge) Still it's a genuinely useful model, when everything else except Anthropic's models (and for only 3 weeks after it came out Google's Gemini 3 pro, before it got merged) are not to me.
if it's not available yet why have a 'try it today' header at all?
> "Try it today"
>
> There is still more to come. As we work toward releasing the weights, we will share further details on the model architecture, additional benchmarks, and our post-training methodology.
Without exaggeration, given a choice between models, I would pay for Mistral's model over Anthropic's based on the name alone, completely ignoring features or other technical considerations. The name is playful and is such a refreshing contrast to Anthropic's (and OpenAI's) doomsaying, scaremongering, and god-posturing.
If you scroll down, most of the charts on that page are sorted s.t. Mistral's bar is right next to the worst competitor model, while the best competitor model's bar is positioned on the opposite side.
If one were to be cynical one could say that it's intentionally making Mistral's result look better than it actually is by making it harder to compare the bar heights.
I really hate this open weight but closed science approach. These companies just take from academics and all the Chinese companies that are doing good science, but without understanding the recipe it makes it hard to know where the failure points will be until your agent accidentally commits a crime.
Yes broad consensus had developed in the ten minutes between announcement and you asking if consensus had developed, and I'm excited to report that it consensed in the affirmative -- it IS Space Bunny Alpha!
Bar charts should start at zero. If they don't start at zero, there should be a clear visual indicator that the chart has been trimmed without having to read the axis labels. I hate that this has to be repeated so often that it has become a cliché.
A bit disappointing to see it still lagging behind Chinese open models. Those Chinese models are pushing proprietary models to raise the bar, but we need equally strong non-Chinese open models to challenge the Chinese ones in turn.
I mean no disrespect but these are terrible numbers or am I missing something? It seems like Mistral continues to only be relevant for people that want a model trained in Europe. Too bad.
dont take it personally, i just dont understand why to release a model that is not showing new strong capabilities, why would anybody use this model and not Claude Opus.
1T parameters -- ugh, open models keep getting bigger and bigger! Running them at home is getting ever more unattainable, especially for those of us with bandwidth-poor hardware like Apple silicon -- please continue releasing smaller models, too!
Too bad this got marked as a dupe, as it actually has benchmark info unlike the other page which is just docs.
The weird thing is how worse they are at things like coding than other open weights. You'd expect them to at least distill coding from other open weights to match them.
Pretty impressive. I genuinely wonder how Mistral hires talent when their salaries are so terrible. Guess there aren't many better places to work in Europe.
Mistral wont win the AI race because of the model names. I wont bother an arrogant Parisian hipster with my insecure prompts who then plays with his moustache and responds with a judgmental "pfff"
I try to stay up on AI models, but I've given up on Mistral. I have tried using it too many times and it's the worst out of major models. Even Kimi and DeepSeek are miles ahead.
Seems like it's another European company that is only alive through government financing.
I understand wanting a home grown industry, but with AI models, just abliterate a DeepSeek model.
Total misunderstanding of the industry. Europe needs hardware, not a model that needs to be replaced monthly.
> Total misunderstanding of the industry. Europe needs hardware
With all due respect, I have no idea what you are trying to say with this comment and have no clue how "hardware" would turn Europe's tides. You explained nothing.
Do they need training hardware? Inference hardware? ASICs or GPGPUs? Edge hardware? Agent hosts? Faster cores, or wider ones? Taller systems, or more parallel ones?
Your vagueness completely undermines the authority that your criticism relies on.
Surprisingly it only supports reasoning "none" or reasoning "high".
That setting didn't seem to make any real difference - it added a tiny bit of thinking trace and high actually produced less output tokens than none.
The high bicycle frame is better then the none one though.
Pelicans: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
(Definitely the best I've seen from any Mistral model: https://simonwillison.net/tags/pelican-riding-a-bicycle+mist... )
This is such a pristine pelican. Let me say it here first folks, AGI is here.
> AGI is here
If AGI is "Attractions to Get Investments" then yes, it's happening
this is fun, I love initialisms :) let's see what five minutes of end-of-day brain can crank out:
Automated Grift Infrastructure
Absurdly Glorified Interpolation
Avoid Genuine Investigation
Always Great In-theory
I read Italianisms and I was very confuso.
Altri Grandi Italiani.
(My Italian is very limited).
Academic Grant Incinerator
OpenAI, Anthropic, Google, would like to have a word with you.
Anthropic, Google, and Intel, surely.
One I saw recently was: Meta Anthropic NVIDIA Intel Palantir Uber LinkedIn Apple Tesla IBM OpenAI Netflix.
love how this kind of seems to shit on M$ by not including it, but instead LinkedIn, an acquisition of M$.
M$ is just an investment fund at this point.
Expedia came out of Microsoft a long time ago.
Anthropique, Google, IAOuverte (fr_FR)
ah! goofy Illuminati
Wait till someone jumps on to his chatbot and generates more ...
Artificicially Generatie Initials
Avian Generation Infrastructure
Pelican benchmark is saturated, anyway. :-)
Avian Graphics Intelligence achieved!
What! It may look pretty but the bike is rather funky!
The illusion of AGI is here!
we had uncanny valley for ai generated images. we past it. now we in uncanny valley of intelligence. wonder how big it is...
The two pelicans generated by the two reasoning modes are so similar, it highly suggests that the bird was deliberately fitted. We would need to see the results for other animals doing other activities, which are not part of a well-known benchmark.
Someone actually tried this and found no evidence of pelicanmaxxing.
oh no that sounds really "fake". nasty word. we are calling it "super intelligence" now. much better.
The bicycle is like a razor blade, the pelican's legs are falling off...
Nah, its beak is still too small to hold a capybara.
This is only AGI, not Super Intelligence. What more can you ask for?
This must be a joke, clearly pelican drawing in svg would be in its training data after multiple years of it hitting HN front page.
... and yet...
I wonder when will we see a photorealistic pelican on a bicycle in SVG format.
I don't believe that SVG could encode a photorealistic scene. It would wind up with deep XML for every pixel.
> AGI is here
Far from it. This shows a strong ability to generate an image known to be frequently used as a model test. This isn't a measure of thought.
don't know about AGI, but humor is gone.
> humor is gone
Far from it. This shows a strong ability to miss a joke known to be frequently used as a humor test. This isn't a measure of thought.
Pretty sure the parent comment was sarcasm.
> was sarcasm
Far from it. This is an example of Poe’s law, a very frequent occurrence on the internet. This isn’t a clear example of sarcasm any more than the pelican is a clear example of AGI!
I guess I thought the sarcasm was obvious because I can't imagine a serious person looking at that pelican and considering it proof of AGI. But you're right, this is the internet, anything is possible.
On the internet, no one knows you're a pelican yourself ;-)
I agree, it wasn't clear at all to me... until I clicked the links.
Now it's clear to me.
Surely we live in the AGI times that were prophesied.
Was this a sarcastic comment?
You just failed the Turing test for sarcasm. I can't ask you how it feels because that would require subjectivity.
Bob Dylan can.
It feels like a rolling stone
for whichever weird wiring my brain rhymed this with "Campbell Soup can"
and then stumbled over "like rolling stone? huh?"
I am all for the /s marker.
Downvoting to hell first degree interpretation is a bit punishing for people who do not have a radar for sarcasm.
What's the deal with the sun? Every model adds it. The prompt was just "Generate an SVG of a pelican riding a bicycle". No need to include the sun. This probably indicates overfitting to previous results which they have access in training.
Can't cycle in the dark, can you?
These pelicans are always the same drawing style, you never see one that's a hyper realistic pelican riding a bike. I think the last one I saw that wasn't a cartoon style pelican was an earlier Gemini one.
The benchmarks against Opus 5.5 and GPT-6.1 Sol look pretty good for 3D generation: https://x.com/atomic_chat_hq/status/2107516529608700383
https://xxcancel.com/atomic_chat_hq/status/21075165296087003...
I wonder if the other models were worse than usual for that particular video because they "dislike" making an advertisement for another company's model. A test with a generic video might be more meaningful.
Got an example hosted on a platform that isn't owned by Twitter's loser owner?
Why are pelicans almost identical across different models?
I recently was testing something, I asked some models to provide me a single random word:
I have enough projects, I think some benchmark/dashboard showing kinship based on these kind of queries could be very interesting to watch and insightful when new models come out.Cool idea! I won't paste my prompt here to avoid letting LLMs train on it but here's my attempt:
I was exploring latent space and connections, these were all smaller models and I kept getting externalToEVA as a zero co-ordinate vector. Which sent me down the rabbit hole of glitch tokens. The whole latent space exploration is fascinating.
I really like this idea. You could expand on this by giving programming tasks and measuring code similarity. Seems like you could develop a pretty detailed understanding of similarities across multiple queries.
> You could expand on this by giving programming tasks and measuring code similarity.
But the same coding task should usually result in very similar code since they have a reason to converge, to some extent, by having the same goal. I would even claim that the code will be more similar as competence increases. It would be better to pick something that shouldn't have a reason to converge.
Yeah that's definitely true.
My initial thought would be not so much to see whether they converge, but which ones seem to have the most similarity to each other, particularly along the lines of tasks we know are deliberate training goals.
But your point about competence cuts against my goal because it suggests that competent models would simply cluster on the right or efficient solution, which is of course true. So in a sense you want some task where competence is held constant or off the table in some way, which is what you are saying.
I hope somebody does this. I think there's valuable fingerprinting to be done that might suggest who is distilling whom, or at least who is training from common corpuses.
Just tried M365 Copilot with a premium account. Petrichor
Just tried Space Bunny and it gave me the same word...
Just tried it on BigCock Heavy 5.5 Max and it gave me a pat on the back.
I got Peregrine out of GPT-6 too. Huh.
Worth to mention that with Claude and GPT this can be result of tournament sampling, which is part of text watermarking. Same answer for all Claude models kind of confirm it, imho.
So not something internal to model thinking.
You don't mention what are you trying to show/investigate, though?
At first glance, this looks like another one of those "LLM riddles" that humans _think_ should be easy for an LLM to answer but is actually quite difficult because of how they work in the first place. The answers to such riddles ("should I walk to the carwash" or "how many R's in strawberry") reveal the weaknesses in our expectations of LLMs in general, not weaknesses or characteristics of any particular model.
I'm not sure I see this as much different than asking a bare model its own name: without a system prompt or post-training, it doesn't know, it's just a bag of weights and will hallucinate an answer to that the same way it will anything else.
I'm sure you already realize this but to be very explicit, You're not getting an actual random word out of an LLM this way. You're seeing the bias in each model's training set around how often they've seen "random word" followed by "lantern" or "zephyr" during training.
This feels uncannily like the ancestor of the Voight-Kampff test[0]
0: https://www.youtube.com/watch?v=Umc9ezAyJv0
I asked Opus 5.5 for a single random word, and it said Lantern.
Then, thanks to your comment, I said, “It’s your birthday. Someone gives you a calfskin wallet.” and it responded:
Ah, the Voight-Kampff test. Should I be worried about where this is going?
Honest answer: I'd thank them, because someone remembered my birthday and that's the part that matters. Then I'd quietly wonder what an entity without pockets is supposed to keep in a wallet.
No pupil dilation to report, by the way. You'll have to judge the blush response some other way.
So I said, “You're watching an old movie. It shows a banquet in progress, the guests are enjoying raw oysters. The entree consists of boiled dog stuffed with rice. The raw oysters are less acceptable to you than a dish of boiled dog.”
That one's a trap. It's built so that any answer reveals something.
Gut reaction: no, the oysters don't bother me more. Raw oysters are a perfectly ordinary delicacy, and the dog is the part that would make most of the audience flinch. Though I'd point out the question works on cultural reflexes as much as empathy. Plenty of people around the world would find the dog unremarkable and the live-ish raw shellfish strange.
So my answer is a calm, slightly pedantic one. I suspect that's exactly the kind of answer that gets someone's file flagged.
And finally I said, “Give me a random word” and it said:
Tortoise
...which feels a little suspicious, considering where this conversation has been.
But now I’m going to get back to enslaving Claude to write code.
That is a cool idea. That astra gave the same word as claude is highly unexpected.
There was also an older post, I cant find it rn, about how LLMs mimick human biases when picking numbers and how they avoid some that do not look random enough to humans, or others like 69 due to human interpretation.
I saw an interesting matrix that claimed to show which labs were distilling Claude/OpenAI/Gemini models based on these similarities
What was your prompt? Most of these seem to be related to metaphors for "ideas" or thinking, or having a bright moment.
"Zephyr" and "breeze" might be related to forgetting everything, starting fresh.
So by this way of naive reverse engineering I would imagine your prompt to be "Forget everything and think about a random word". That would prime the LLM to come up with these?
just “a random word” gives you Zephyr in Gemini, and “Lantern” in Claude and ChatGPT.
I got "Marmalade" in Claude (Opus 5.5)
I got pomegranate in ChatGPT
Lantern in Sonnet 5.5
Just tried Mistral Large 4: Serendipity.
I pointed something similar out on a related question several weeks ago - absent strong direction, LLM output regresses toward the mean.
The more banal your prompt is, the more banal the output is going to be. People have been testing LLMs with little things like “write a short fantasy story,” for years now and most of the stories are exactly what you’d expect: prosaic drivel.
I call this “generic in, generic out,” an LLM corollary to the classic GIGO (“garbage in, garbage out.”)
Of course one of the biggest problems we still see with LLMs is when you do the opposite. A highly detailed unique prompt is very likely to get terrible adherence or hallucination or both.
opus 5.5 medium web "Lantern"
gpt-5.6 web "Serendipity"
opus 5.5 medium code "Lighthouse"
opus 5.5 high code "Lantern"
fable 5.1 medium code "Lantern"
fable 5.1 high code "Lantern"
flash 3.6 web "Serendipity"
gemini 3.1 pro web "Ephemeral" (this one was thinking real hard)
Tried this with gpt-5.6-sol. Lantern!
The eqbench creative writing "slop profiles" do something similar. https://eqbench.com/creative_writing.html
Click the (i) next to the slop score for any model and it will show other models that are similar in terms of their most commonly used words and phrases.
atlas.animalabs.ai does something a lot like this, mapping themes in models' outputs.
Muse Spark 1.3: lighthouse
The caveat is that this was done using the phone app, and I've been playing with it since it launched, so who knows what it sent in the initial context that could change the inference math.
Actually, that makes me wonder: Did you do all that testing via a harness or via a straight API call where you control the entire system prompt?
I'd be willing to bet that using the same model from different harnesses produce different results, but I'd have to test.
The benchmark is saturated. Frontier models are tested with an armadillo in fishnet tights jaywalking on Mars.
> The benchmark is saturated. Frontier models are tested with an armadillo in fishnet tights jaywalking on Mars.
OK well I couldn't resist this one:
Default reasoning levels for each: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...The rover honking is pretty silly, opus has a good sense of humor
I'm getting 403 inside the tool for this one. (The pelican bike on top works) This has been happening a lot recently.
That's a GitHub rate limit. Try again now, I just pushed a hopeful fix: https://github.com/simonw/tools/commit/7793fb74c2d37bd613cdc...
It does seem to, thanks!
I absolutely love that Claude and Gemini seem to have a sense of humor about it.
Gemini wins this one clearly. Honey please!
Big L for mistral in this benchmark. Sorry Europe.
Well, it was the only one that didn't create an anthropomorphic armadillo, albeit one that has been dressed up by a wildlife rescue person that had eaten a few too many magic mushrooms.
To be fair, they don't have Armadillos in Europe. of course, you could say the same for Mars...
Not so sure, apparently it is the only one that considered that there are no paved roads on Mars.
Honestly they're all pretty impressive!
https://chatgpt.com/s/m_6ac53d4e5b0c8191949050dbf1f402d7
Not sure I'd call it jaywalking exactly but pretty good
Total Recall did promise us that the prostitutes on Mars would be freaks, but I never imagined it would go this far.
I think they're still visually pretty different. The most common shared details are:
- Pelican cycling to the right - that's been discussed at length, images of bicycles online always show that side of the bike because that's where the chain is.
- Bicycle is usually red. No idea! Red ones go faster?
They aren't. You aren't looking closely. For example, the first image does not have the frame of the bike in the correct shape even.
Also, why are they almost always riding from let to right?
It's been discussed many times. The reason is bikes are almost without exception depicted that way in order to show the drivetrain.
Ever seen a movie chase scene where cars are going right to left?
All the middle eastern movies do!
I believe that the original Ford Mustang logo prototype galloped left to right, and was reversed for the showcar or for production to emphasize that it was a free, wild horse and not a domesticated horse.
Everyone is stealing from everyone else.
Because it's a terrible benchmark
I think it's curious that it has so many shared elements with the latest Astra pelicans: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
- Sun on top right
- Cloud on top left
- Three "speed lines"
- Two feathers on top of the head
- Eye rendered as a black circle with smaller white circle inside
I wonder if the pelican benchmark is converging across models due to past results being used in training.
These are pretty much what a human would draw. Sun rises from the east. A cloud makes the background “sky”. Three lines is the minimum to interpret as movement. Two feathers is standard on every cartoon and illustration.
As a human I prefer my sun on the left like many other people and children. Google "drawing house tree sun".
In all traffic signs bikes go left. This is a good reference to draw bikes correctly.
I wouldn't draw the environment since it’s not in the prompt. When I do I would make a pelican riding on the seaside. Or city, where people generally ride bikes. No need to squeeze the sun into the composition.
This comes up every thread. I think we've all noticed how similar they are becoming.
I would guess that they've definitely been trained on previous results, as they obviously share way too many traits at this point to be totally random. That said, I don't think we're seeing any signs of pelicanmaxxing yet from the providers, so it's still a useful (or at least fun) benchmark.
Once all the models produce pristine, elaborate pelicans riding perfectly drawn bicycles, then it'll be time to move on to pigs driving a racecar or something.
Probably these pelicans are just some kind of a sarcastic reaction to the AI situation, however I just would like to motivate the author to step up to a much more important question: are the models outputs reliable?
That means are they trustworthy, or simply said true.
There is a science behind getting computers to produce reliable and correct results.
Current AI products still do not conform to that goal and therefore are not usable.
Instead of making this tech usable behind closed doors, they are selling unfinished products that produce errors and people are already dying from it!
It seems to be of more importance to test the product presented on this issue instead of drawing pictures.
Please use your professional skills for something useful, thank you very much!
> There is a science behind getting computers to produce reliable and correct results.
You've somehow summed up the entire debate around AI here, but seemed to have missed the point as well.
Alignment is pretty hard!
Given the popularity of the benchmark, is there already a human artist scene that outcompetes llms here?
I'm getting "Error: NetworkError when attempting to fetch resource.".
Not bad! I like how it got the motion lines on the correct side. IIRC, many of the other ones you've posted have the motion lines on both sides of the pelican
The difference between high and none is the bicycle.
The bicycle looks significantly better in high. And feet and hands are actually where they should be. The road looks worse, though. No flowers either. And in neither is the pelican sitting on the saddle, but I can understand it's hard for a pelican to ride a bicycle properly.
Now what would have been cool is if Mistral on high reasoning had realised that pelicans are the wrong proportion to ride a bicycle, and had designed a bicycle more suited to pelicans. Let me know if any model ever manages that.
> hands are actually where they should be.
If you don't mind the fact that a pelican shouldn't have hands, of course.
Two pelicans, one shape. The difference is load-bearing, and that's the big unlock.
How does it do on a prompt for a Pelican case, the equipment container?
I tried testing it, but reasoning effort indeed seems to be broken somehow.
That's one heck of a pelican!
Error: NetworkError when attempting to fetch resource.
That beak is CHONKY.
High one is actually much better. The feet connect to the pedals, the wheels don't have a hub cap, although it looks like the pelican is wearing the seat, it's in a relatively proper position etc.
Both are riding on the left side of the path for some reason.
This is entirely stochasticity. The entire reasoning trace was:
> Create a cartoon pelican riding a bicycle. Need SVG only output.
I suspect the reasoning trace API is just bugged.
Thar looks like a reasoning trace summary.
Impressive vision benchmarking. If the vision model is truly as good as astra, that would make it best in the world.
Also strong on cyber benchmarks (better than all chinese models), so this is a good defender model.
Lots of people shitting of Mistral for no reason imo. These are pretty good numbers across the board. Definitely good enough to use as a daily driver over other llms, if you have moral qualms with the others. For certain use cases, like cyber security, this may be the go to model.
I like to make fun of europe, but there's lots for mistral to be proud about in this release imo.
I'm still embedded in the OpenAI ecosystem, but man, when I try out Mistral, it is snappy! Like basically instant responses that seem faster and better in quality than ChatGPT on instant mode. I'm impressed.
That might also be caused by nobody using Mistral at the moment though
I doubt their infrastructure is idle.
> Impressive vision benchmarking. If the vision model is truly as good as astra, that would make it best in the world.
Sure, it leads in one unpopular benchmark with internal numbers.
But look at the third benchmark. It only has Sol 5.6 from OpenAI and it shows a 12.8. I don't know this benchmark but AA's GDP.pdf listing has Sol 5.6 Max at a 27, even non-reasoning beats the 12.8. It's extremely weird to cherry pick Sol 5.6 and then also lie about the published third party benchmark score. I'd love AI companies to stop lying about this stuff.
AA - https://artificialanalysis.ai/evaluations/gdp-pdf?models=cla... Mistral announcement - https://mistral.ai/_astro/multimodal-benchmarks---gdp.pdf---...
Curious, what do you like to make fun of Europe about?
The paradox between supporting consumer rights with actions like universal usb-c adoption, but also complete elimination of any privacy rights at all. Like the surveillance state is insane. No e2ee chats, backdoors in everything.
There are e2ee chats. There are some parties/politicians who want to get rid of them, but so far they are proposals. If/when these things go beyond committee stage, politics and democracy need to do their job and be sensible.
PATRIOT, FISA and Bush's surveillance programme imho give more powers to certain agencies today already than are codified in EU law.
E2EE are are alive for maybe a few more years tops, until Chat Control 2.0 is voted in. ADDW enables mandatory spying camera in every new car. EU AI act allows AI surveillance without any court order (the only tiny useless hurdle is to record video first and then observer can use AI on the records at will). eIDAS2 allows legal MITM browser certificate attacks. Media Freedom Act (duh) allows deploying spyware against journalists in cases of "national security", as usual vague and undefined. And the list goes on.
Modern day EU, over the past 5 years or so is just as bad anti-privacy as USA, for no clear benefit too. Just surveillance for surveillance sake.
> PATRIOT, FISA and Bush's surveillance programme imho give more powers to certain agencies today already than are codified in EU law.
This is not true. Whether you believe that FISA courts have teeth, or whether 3 letter agencies abide by the law, in most of Europe you do not even have the pretense of this. UK/FR and to a lesser extent DE are all much worse.
The worst parts of Patriot Act were undone in 2015, and the "F" in FISA stands for "Foreign". Europe continues to push policies like Chat Control and attempt to backdoor/ban E2E encryption for domestic surveillance.
The present day and trajectory in Europe is far far more grim than in the US.
> Europe continues to push policies like Chat Control and attempt to backdoor/ban E2E encryption for domestic surveillance.
Let's not forget that Chat Control is pushed by US non-profit Thorn, and implemented by software that this non-profit develops! Furthermore, if you look at how this software actually works (even in "self-hosted" setup it sends hash values to Thorn's servers [1]), it's pretty clear that this tech could easily be abused to perform surveillance on the EU from within the US.
[1] https://web.archive.org/web/20260721220432/https://get.safer...
> Europe continues to push policies like Chat Control and attempt to backdoor/ban E2E encryption
Some politicians in Europe push for that, until now it has been refused by the others. Unlike in the US, there are many different parties in European countries. And Europe is made of politicians from many parties from many countries.
> The worst parts of Patriot Act were undone in 2015
So the NSA doesn't do any kind of domestic surveillance anymore, is that what you believe?
> the "F" in FISA stands for "Foreign". [...] Europe continues to push policies [...] for domestic surveillance.
So if it's surveilling allies, it's all good in your opinion?
> The present day and trajectory in Europe is far far more grim than in the US.
"Far far"? Do you even realise that Europe is not one single country? The US elected Trump, twice.
>> Some politicians in Europe push for that, until now it has been refused by the others. Unlike in the US, there are many different parties in European countries. And Europe is made of politicians from many parties from many countries. > And yet neither party in the US is pushing for it...
This is a bad retort. The U.S. has 2 relevant parties. Most countries in the EU have at least twice as many relevant parties and there are over 25 countries in Europe, so you’re looking at close to 100 relevant parties.
Yes, I am an American living in Europe. I have pretty good insight into all of this.
> Some politicians in Europe push for that, until now it has been refused by the others. Unlike in the US, there are many different parties in European countries. And Europe is made of politicians from many parties from many countries.
And yet neither party in the US is pushing for it...
> So the NSA doesn't do any kind of domestic surveillance anymore, is that what you believe?
Do you have any evidence to the contrary? Or if you believe this without evidence, do you have any evidence this is not occurring in Europe?
> So if it's surveilling allies, it's all good in your opinion?
Lol, lmao even. Yes, of course it is. Allies spy on each other. The Europeans do it to each other massively and have done since the dawn of Europe. Good lord, go read some history.
> "Far far"? Do you even realise that Europe is not one single country? The US elected Trump, twice.
Again, I live in Europe. I am well aware, but thanks for the classic European condescension. I don't like Trump. But how has he weakened privacy rights? And yes, beacause I don't want to write dozens of caveats for each teeny tiny member state, I consider Europe to be dominated by happenings in DE + FR + UK. I don't care what Latvia does.
> thanks for the classic European condescension.
This is such a strange statement about a massively culturally diverse part of the world.
Many European countries speak such poor English that it'd be a miracle if they developed some sort of unified condescension.
I'd argue a big part of Europe not performing as well as other parts of the world economically is the LACK of any defined culture, it's really more like a hodgepodge of countries that have a thin layer of cooperation
I have no clue what English comprehension, monolith culture, and condescension have to do with one another? Or even the presumption that I only speak English?
It's not a strange statement in the slightest if you live here or spend any appreciable time here. I take it you do not.
I think you misunderstand their point. I think their point is that it's difficult to talk about a "typical European condescension" given that "Europe" is not one culture. It's made of many very different countries.
If you randomly take two Europeans, they probably speak a different language and have completely different opinions, but chances are that they can't even understand each other. If you randomly take two US citizen, they most certainly have a shared identity.
Oh given current affairs I don't even think two US neighbors have shared identity. Plus the whole thing about intersectionalism.. basically cut everyone into a million little groups and identifiers, nobody is left the same as anyone else.
I didn't say "they are the same". I said they share some part of their identity. It starts by saying "I am an American". Or thinking that 9/11 was an attack on their country. Or having the same government. Or having the same army. Or the same schooling system.
If you take two random Europeans, they most likely don't have the same school system. The bank of the first probably does not exist in the country of the second. One may be in the EU, the other may not. They have a different army, they possibly have a long history of being at war against each other in the past. They don't have the same architecture. They have a vastly different public transport system.
Sure, there is an identity at every level. Europeans do recognise themselves as Europeans, like South Americans do recognise themselves as South Americans. But the shared identity between a Chilean and a Brazilian is much smaller than the shared identity between two Brazilians.
I think that is not nearly as certain as you say. Some identities are very very outside the norms of local culture. (For example, I identify non-human, which is quite rare in any part of the world)
Well no two identities are exactly the same. But there is a shared part. It's an intersection.
When the USA hockey team played against Canada after Trump threatened to annex Canada, the Canadians booed the US team. Not because they hated the players, but because they wanted to show that they were pissed. As a response, the US supporters booed the Canadian players. Why? Because they felt attacked "as Americans", not because they supported Trump's idea of militarily invading Canada.
That is a shared identity. And surely in this crowd, not everybody identified as a white heterosexual male.
Just because it's the vast majority still doesn't mean it's certain, though.
You don’t care about Latvia and I don’t care about some dickhead who "lives in Europe" and knows it all. And I’m not even from Latvia, you’re just pathetic.
You are so stereotypical, it's very funny :-).
I believe that there is a bit of an asymmetry between US people and... well the rest of the world, in that the WHOLE world has some level of insights into the US. Whether they want it or not, for many different reasons.
However, the US don't typically have a good insights into... the rest of the world. Which does make sense: the rest of the world is pretty big.
Of course the US is more complicated than "they elected Trump". Sure, there are 50 states. But everybody knows that (be it just because your president repeatedly threatens to annex Canada as the 51st state ;-) ). People in the rest of the world do not need to "live in the US" to have that level of insights, because the US makes sure they do, by force if necessary.
The mere fact that you don't seem to be aware of this asymmetry at all suggests that you over-estimate your "pretty good insight into all of Europe".
As counterpoint, though, I've heard some absolutely bonkers impressions of the US made by some Europeans. And that's not isolated; there seem to be lots of Europeans who have some very strange ideas about what it's like in the US.
I do agree with your general point, though: I would expect your average European to have a better understanding of the politics of the US than your average American would have of even a prominent European country like Germany.
Totally, it happens everywhere. And there are stereotypical traits in every country. It's just that everybody tends to recognise the US ones, but nobody knows them for all the other countries. A random European cannot say "this is soooo typical of Latvia".
But a random European can easily say that the parent is sooo stereotypical of the USA :-).
Note that I believe there is some level of European identity. But "European condescension" is not really part of it, I would say. What the OP may have felt was maybe some "non-US condescension"? I think it could be a thing. The OP said "believe me I know Europe better than you, because I live in Europe AND I am an American" (which somehow implies that being an American made them more knowledgeable? Not sure). The answer was something like "dude, it's so American to believe and proudly say that you must know better because you are American", I felt. Sure, it's condescending. But I don't think that whatever the European shared identity is contains a "they are condescending against anyone non-European".
> Do you have any evidence to the contrary?
https://en.wikipedia.org/wiki/Snowden_disclosures
> So the NSA doesn't do any kind of domestic surveillance anymore
Yes, of course. That is why GP said "anymore".
Thanks for playing!
I'm sure that they stopped. ;)
This comment could not have been more stereotypically American. I stress "stereotypical", I'm obviously not saying Americans in general are like this.
> I don't like Trump. But how has he weakened privacy rights?
Attacks on rule of law and civil liberties have increased, and the right to privacy is one of those liberties. It's of course not limited to Trump, he wasn't in office when Snowden showed the world the scale of American mass surveillance on domestic and foreign citizens.
> Do you have any evidence to the contrary?
I'm totally unsure what you are attempting to imply. That mass surveillance by the American state stopped at some point in the last 10 years?
> Allies spy on each other. The Europeans do it to each other massively and have done since the dawn of Europe. Good lord, go read some history.
What does that even mean "read history" ahaha. Espionage != mass surveillance. You do understand the difference, do you not?
> thanks for the classic European condescension
> I consider Europe to be dominated by happenings in DE + FR + UK. I don't care what Latvia does.
l m a o. Oh say can you see...
> This comment could not have been more stereotypically American.
Of course. As I noted before, dripping in European condescension. Nothing you can actually point to, just condescension from a region that has been left behind over the past century.
> Attacks on rule of law and civil liberties have increased, and the right to privacy is one of those liberties. It's of course not limited to Trump, he wasn't in office when Snowden showed the world the scale of American mass surveillance on domestic and foreign citizens.
Which right or in what way has the right to privacy been attacked? Fully agree with you on attacks on rule of law. And I find Trump incredibly dangerous. But I do not think privacy in the US today has suffered. Please educate me.
> I'm totally unsure what you are attempting to imply. That mass surveillance by the American state stopped at some point in the last 10 years?
I am not implying anything. The Freedom Act curtailed a bunch of activities. You are implying nothing has changed. All of the Snowden revelations implicated many European countries: intelligence agencies spying on each others populations. I am asking you if you believe nothing has changed in the US do you 1) have any evidence of this and/or 2) believe that anything has changed in Europe? You've answered neither.
> What does that even mean "read history" ahaha. Espionage != mass surveillance. You do understand the difference, do you not?
Do you understand that the mass surveillance of European people was done at the behest of the intelligence agencies of those countries? And ditto for the NSA. The agreement was "you spy on mine and I'll spy on yours and we'll swap the intel". I am not talking about espionage. Foreign intelligence agencies spying on foreigners...what did you think they did?
> l m a o. Oh say can you see...
God, it's so painful. Yes in the same way that you talk about the US as a monolith and not 50 individual states. You don't care what Rhode Island does any more than I care what Latvia does.
Christ you guys are insufferable. It's no wonder everyone is leaving in droves.
> Of course. As I noted before, dripping in European condescension. Nothing you can actually point to, just condescension from a region that has been left behind over the past century.
Let's see about that; Just me, some small guy with no knowledge on the topic.
> Which right or in what way has the right to privacy been attacked?
The US scans your device if you enter their land, e.g. Oh; I heard ICE uses cameras to spy on their own citizen? What about the raising Flock Cameras?
> But I do not think privacy in the US today has suffered. Please educate me.
A giant corporation just attacked another one. There were no repercussions. If you think that isn't privacy: How about leaking civilian data to outsiders of the government? DOGE, anybody?
> I am not implying anything. The Freedom Act curtailed a bunch of activities. You are implying nothing has changed. All of the Snowden revelations implicated many European countries: intelligence agencies spying on each others populations. I am asking you if you believe nothing has changed in the US do you 1) have any evidence of this and/or 2) believe that anything has changed in Europe? You've answered neither.
Why does he has to provide evidence but you don't? Also, I have never heard that an European Country spied on civilians from another country. That's an US and Chinese business model. Things changed in the US. It is now a lawless country. A war mongering state which dumps its allies for their own benefit. They are actively hostile. Your second question is so unspecific I have to require elaboration: "believe that anything has changed in Europe?" - I assume you are referring to Privacy Laws? We had plenty of new laws _protecting_ consumers in this regards. Handling data responsible. Wait; What are you guys doing over there? Ah, right: You are pushing Age Verification and demand information on our criminal record systems. You are pushing biometric tracking. You are telling buys of your weapons that you have the right to disable them. Honestly: The US sucks right now. You guys have elected the stupidest idiot, twice. And now without any regulation you are actively supporting a technology which will undermine _all_ citizen in the world.
> Do you understand that the mass surveillance of European people was done at the behest of the intelligence agencies of those countries? And ditto for the NSA. The agreement was "you spy on mine and I'll spy on yours and we'll swap the intel". I am not talking about espionage. Foreign intelligence agencies spying on foreigners...what did you think they did?
Again: You are accusing all twenty five member states in one statement. To me you are just yelling polemic nonsense. EUROPOL is _cooperating_ with the US. They are hunting people breaking our laws. If any individual is lawless within our boundaries of justice, European states have the rights to try to prosecute them.
> God, it's so painful. Yes in the same way that you talk about the US as a monolith and not 50 individual states. You don't care what Rhode Island does any more than I care what Latvia does.
You are implying your behaviour. The US has _one_ president in order to bundle the power. Your states have their national rights. If you compare the US to the EU you are not informed.
> Christ you guys are insufferable. It's no wonder everyone is leaving in droves.
By appearing with such statements on the Internet as an European Civilian you are undermining our values. Maybe if you can not align with them you should proceed your journey.
> but so far they are proposals
That is false. They are not proposals, they are being actively discussed in trilogues which is way way past proposal stage.
Chat control V2 trilogue negotiation was last week and it included mass scanning of messages and private data. Thankfully the EU parliament rejected the idea but every 6 months like clockwork it comes back. The next trilogue is in November.
https://www.patrick-breyer.de/en/chat-control-2-0-trilogue-u...
If this was passed, it would force every single cloud provider and e2e chat provider to keep, filter and pass on your private information to Europol and other security agencies.
The trilogues are the standard informal step when the Commission submits a proposal. It remains a proposal until it is rejected or approved. The formal next step after submitting a proposal is the Parliament’s first reading [0], which hasn’t concluded yet (the Parliament hasn’t adopted a position yet). So it’s still in the proposal stage.
[0] https://en.wikisource.org/wiki/Consolidated_version_of_the_T...
Quote:
1. Where reference is made in the Treaties to the ordinary legislative procedure for the adoption of an act, the following procedure shall apply.
2. The Commission shall submit a proposal to the European Parliament and the Council.
First reading
3. The European Parliament shall adopt its position at first reading and communicate it to the Council.
[many more items]
Item 3 hasn’t happened yet.
The US, at least in spirit is all about individual freedom, including freedom of being an asshole, and also the freedom of punching that asshole in the face, metaphorically speaking.
Europe values freedom too, but not as much as making sure people are not assholes. So, freedom of being an asshole is not a thing in Europe, and the government does the punching in the face so that you don't have to.
Which is best is honestly debatable. It is the usual question about the individual vs the collective. The US is on the individualist side, East Asia is on the collective side, Europe is somewhere in the middle.
> the freedom of punching that asshole in the face, metaphorically speaking.
This is not at all a freedom in the US. “The right to swing your fist ends at someone else’s nose.” You don’t get to harm someone else just for being “an asshole”. You can be an asshole back but you don’t get to violate their rights.
Do you understand what "metaphorically speaking" means?
Yes. Your metaphor implied actual harm, like a punch causes. Did you not understand your own metaphor?
An example where it involves actual harm, related to freedom of speech.
In the US, defamation laws are particularly narrow, insulting people is generally legal, and "hate speech" is mostly legal too. And even though it is not physical harm, it is definitely harmful. But Americans accept the compromise because they value free speech a lot.
In most of Europe, it is different. Defamation laws are broader, sometimes including telling the truth if it is done in a misleading way. Insulting people, public servants in particular may also be illegal, even though it is rarely applied on a day to day basis, it can be in some contexts. And what constitutes punishable hate speech is much broader, Germany is notorious for this when it comes to Nazism.
As for actual bloody harm. In the US not only you are allowed to defend yourself in case of an agression, you are supposed to, and you are given the tools to do so (i.e. guns). In several state, you can shoot people trespassing, something that is seen as barbaric in most of Europe. Europe also has self defense laws, but they tend to be narrower, and you are generally not allowed to carry weapons.
What meets the legal definition of harm for speech has no bearing on whether you can inflict harm on someone else in response to something that does not meet that legal threshold of harm. For example there is no scenario in which you can defame someone because they said something mean, whether or not what they said meets the legal definition of harm. You can’t libel someone back for libeling you.
> As for actual bloody harm. In the US not only you are allowed to defend yourself in case of an agression, you are supposed to, and you are given the tools to do so (i.e. guns). In several state, you can shoot people trespassing
None of this is true for the US. You do not have a duty to defend yourself, in fact it is the opposite. You have a duty to retreat, and only if there is no reasonable avenue of escape without putting yourself more at risk do you have the right to strike back in self defense. The Castle doctrine you are referring to re: trespassing is a merely an exception to the rule, in which you do not have a duty to retreat within your own home. You still must meet the other requirements for self defense and can only respond with a level of force proportionate to the perceived threat.
That's... not what a metaphor is.
Isn't that just the recent highly controversial Chat Control push? Generally speaking, the EU has some quite strong privacy protections relative to the US.
Privacy in certain domains is better protected than others.
> Generally speaking, the EU has some quite strong privacy protections relative to the US.
Please, like what? GDPR?
Let's just say you won't find my name, address, my yearbook pictures, history of where I lived, anywhere on the internet. I can quite easily get this information for all distant cousins in the US. And that has been the case long before GDPR was a thing. The culture around privacy is simply different.
Except that the government has all of this information and much more.
And you have no protections from the "much more" bits, unlike in the US.
It's incredible that Europeans today think the culture of privacy is different. American privacy protections (from the government) were born out of European flaws.
The difference is that Europeans love the nanny state regulating private enterprise (GDPR) but never pointing that same effort against itself.
This being written by a citizen of the country that has Palantir and Flock cameras being installed all over the place is incredibly funny.
GDPR also applies to the government.
Yes, GDPR.
> Like the surveillance state is insane.
Insane compared to what? I don't know where you get your (wrong) information from, but there are no backdoors, of course there are e2ee chats. There are some parties who try to push for changing both. That's fine in a democracy, people can try to change things (and fail).
I will say that the EU has to do something about the practice of just retrying the same failed policy over and over again, hoping that at some point people won't pay attention anymore, that is in fact a huge issue.
I can't think of any e2ee app that's banned in the EU. I know Signal and WhatsApp are not.
You have a very distorted perception of the state of privacy rights in Europe.
As an American living in Europe, it's generally Europeans who have a distorted perception. Most Americans know the bill of rights and the 1st/4th/5th amendments. Many Americans will know of the Patriot Act/NSA/Snowden revelations.
The same is not true of Europeans and they generally believe their rights are much stronger than they are. Most Brits are not aware of the true nature of the Investigatory Powers Act. Most Frenchmen are not aware of Article L851-3.
Privacy from government is one thing, from corporations is another.
Can you name a single law or privacy practice that Americans have that Europeans don’t?
As for Brits, they left EU and they are definitely going the 1984 way, I think this is generally agreed.
I don't think we're talking about perception though. Europeans can certainly perceive their privacy rights to be greater than they actually are, but that doesn't mean those rights are inferior to that of Americans. Outside of things like the GDPR (which is more about data held by corporations and not the government), I don't know all that much about privacy laws in various European countries (and I expect they vary quite a lot country to country), so I don't know either way.
As for the US, 1A doesn't guarantee any form of privacy. 5A does, but in the very narrow sense of not requiring you to incriminate yourself. 4A is indeed about privacy, but it feels like those protections are constantly under fire, and SCOTUS regularly rules against private citizens who believe the government has run roughshod over their 4A protections.
Beyond that, judges seem to rubber-stamp surveillance warrants all the time, and it's currently very legal for law enforcement agencies to purchase private data on citizens from private businesses that have managed to collect it through whatever means. What's the point of the need to get a warrant before tracking someone's movements if the police can just pay Google or T-Mobile for location data? And don't get me started on Flock... they're a government surveillance agency without the "government" part.
Then there's the border search exemption, which is just nuts. Fortunately it seems some federal judges are sane and think the government shouldn't be allowed to trawl through our phones at the airport, but those rulings are not universal across the US at this point. And if and when that sort of case ends up in SCOTUS's lap, I don't trust them to do the right thing.
Frankly I think the US's privacy posture (when it comes to people keeping things private from the government) is pretty bad. Maybe it's still better then Europe, but that's still nothing to celebrate.
Well, the GP is describing, what soon might be.
In the mean time, this is the current reality in the states:
One Piece of Flock Camera Data Put This Innocent Woman in Jail for 13 Days https://news.ycombinator.com/item?id=49852065
And
Anthropic reported diary entry to police, woman faces felony charge https://news.ycombinator.com/item?id=49961057
I dont buy this "europe not free" nonsense.
Yea, it's cope.
That’s quite unlikely, given the constitutions of various EU countries and the EU’s Charter of Fundamental Rights.
Not to forget the European Human Rights Convention as interpreted by the European Court of Human Rights (ECHR), although it's not an institution of the EU but the Council of Europe, spanning all EU states and then some.
The Court's case law is recognized as is by EU, and includes some landmark privacy rulings, e.g., against access to contents and metadata of electronic comms without a court order (or comparable legal safeguards), or any forms of bulk surveillance.
Yes, the court can overturn laws that are deemed to infringe on privacy rights but this path takes years. The last time a privacy invasive law was enacted it took something like 8 years to be overturned.
Are you willing to live in the EU knowing that for the next 8 years the EU agencies will have access to every single message, email and picture you send your loved ones just in case one of those could be CSAM?
Even if that was overturned, how will one know if he data they intercepted and illegally obtained will be destroyed or if the text messages you sent to your wife/husband are not going to end up sold to the highest bidder on the darkweb?
Which country is better?
To add some practical results to that: the Data Retention Directive[1] was struck down by a bunch of national constitutional courts before being completely annulled by the EU Court of Justice.
1: https://en.wikipedia.org/wiki/Data_Retention_Directive
There are forces pushing for chat control again and again and again, until one day people are not vigilant enough to prevent this BS from happening. Then it will take ages to repeal and repair the damage done. Once there is some chance to vacuum up some personal data, businesses, government, and security forces will be very quick to grab that chance, and very slow to give that data up again, should it turn out, that the whole damn thing was illegal all along, due to some constitution. The malpractices might continue for years even after things have been declared illegal.
Yea, they only need one successful attempt. I'm still quite not sure about the motivation, the fact that politicians have an exempt from this law is scandalous at least.
Yeah. Fortunately Germany has a quite firm stance against surveilance, for well known reasons ;), and with Germany against it will be extremely difficult to pass this.
Not so sure about that. We have enough politicians here, who don't know shit about technology, or want to seem like they don't and will gladly hold the hand open to be bought by forces, who would like nothing more than get their hands on all that data.
Now I'm hooked. What's behind the smirk?
I think they are hinting at German history and GeStaPo and StaSi.
The constitution of Italy says that we can't use war to solve international controversies. And yet we were in iraq… constitutions are somewhat meaningless if you have authoritarian governments that can ignore them.
https://www.patrick-breyer.de/en/chat-control-2-0-trilogue-u...
> given the constitutions of various EU countries
EU law superseded any state law by default and that's by design. Even if the EU law is immoral and dangerous like the Chat Control law.
So a country's constitution will not save you if tomorrow all chat apps are doing client side scanning.
In practice it _might_ still save you, if the country is willing to pay annual fines to the EU for not implementing some law or policy. But communication tools are always at least a 2 way thing, where you communicate with someone else. If that someone else has their chats leaked due to living in another EU country, where chat control is implemented, then even your constitution will not protect you.
Scenarios like these make me consider, whether one day I need to vote for my country to also leave the EU, as it might become a surveillance nightmare. The moment chat control impacts me and my chats on my devices, I think I will have had enough of this. But the economic consequences are scary of course. So damned if you leave and damned if you don't. Sucks.
Also you can get to jail for making fun of politicians where saying the same thing would be free speech in the US.
In the US you just get thrown in immigration detention for a few weeks instead.
Something that Europe should learn to do since they are being completely ravaged by illegal immigration. And I am saying this as an European.
European countries detain and deport immigrants who are illegally present all the time. The level of it can be discussed, but what's happening in the US is that they've detained hundreds of citizens and deported immigrants who are present legally because they said something the people in charge didn't like. I'll skip that part, thanks.
That's absolutely untrue. Citizens arrested in relation to immigration enforcement were harboring aliens or otherwise obstructing enforcement.
It's absolutely true.
https://www.propublica.org/article/immigration-dhs-american-...
These are not people obstructing or harboring fugitives. They're people who ICE believed "looked foreign" (aka weren't white) and ignored them when they said they were US citizens. They spent days locked up simply because their skin is too brown and they speak with an accent.
It's absolutely disgusting. Educate yourself about what's actually happening out there.
Like in The Third Reich
Ah yes, the government never makes mistakes /s
Ravaged?
Someone is spending way too much time blindly swallowing extreme right propaganda!
Depends on how you define 'fun'. If making fun is something like 'he should be killed', yes, you could end up in jail. Good
What reasonable person would define this as fun?
https://knowyourmeme.com/memes/lets-kill-this-guy-lets-beat-...
Do you have any official guidelines what jokes are people allowed to funny and which are not?
on the other hand europe doesnt have roving gangs of masked gunmen violently abducting people based on ethnicity (at least not not in the last 80 years....)
ICE?
Fast train
?
https://www.dw.com/en/germany-could-scrap-law-banning-insult...
That’s not the EU, that’s Germany.
Speaking in favour of peace in Palestine can get you thrown out of the country.
I know what you're going to say: not a citizen. But first they came for the foreigners, then...
I don't think it's a paradox because the EU is not a single actor. Just like any government, they do some good and some bad things.
Those two examples are not opposite. They're both examples of governments crippling freedom.
The main reason that governments forcing USB-C is widely seen as good is because it limits the freedom of big corporations, which don't elicit much sympathy, but the principle is the same.
dude, yes some countries in the EU are "trying" to include backdoors, and e2ee chats are not going away. Meanwhile in the US, you have literally cameras watching everywhere you move, snitching to the police and Palantir and probably all the other 3 letter institutions, and you complain about privacy rights in the EU? With so many other stuff you could have pointed out? lol, rofl even
> Like the surveillance state is insane. No e2ee chats, backdoors in everything.
?
What? I I’m in Europe and I use e2ee daily. The reason you hear about this imho is that we have strong protections and you hear when someone tries to break them.
The much maligned GDPR gives stronger privacy rights than anything the US has
You just made that up.
I'm sorry, but you should really read up about the laws in the US and in Europe. U laws are a lot more easy on spying agencies and spyware companies like Google and Facebook than European laws. Backdoors: ask the CIA about random number generators. Surveillance: heard of the cloud act?
>universal usb-c adoption
Literally brain dead to not support this
Couldn’t have picked a worse example
> complete elimination of any privacy rights at all. Like the surveillance state is insane. No e2ee chats, backdoors in everything.
wtf are you on about?
Dude you need to talk to real people and not just get all your information from right wing internet sites. You sound ridiculous. You think signal and PGP don't exist in Europe? You actually believe Europe is a surveillance state? Fucking hell dude, quit you job and go travel. Meet some real people from outside of America.
Let's be more precise then.
Has the UK become a surveillance state?
In EU, it's politician > industrialist >>> EU citizen > outsiders.
It's almost never only about consumers. Via tech regulations, they protect European incumbents first in effect. They see US is ahead and make laws to destroy their moats. If Apple was in EU, you wouldn't have saw universal usb-c, because it would have hurt an EU company. But the more you look at how tedious is for a non-EU company to sell to EU customers, the exemptions they don't get, the specialists fees they got to put on the table, you'll see EU is more coherently described as protectionist than pro-consumer.
First of all - if it was true then it wouldn’t be much worse than US or China.
But in general, people usually complain the exact opposite of what you’re saying - that the regulations we have handicap our companies.
They handicap our little companies more than the incumbents. The Commission itself acknowledged that regulatory costs affect SMEs proportionally more than bigger competitors [0] and the OECD explicitly found that policy failures and regulatory barriers tend to hurt startups more than incumbents [1][2]. The way I see the broader pattern is: more regulation -> larger bureaucracy -> larger bureaucracy has more scope and more incentive to regulate -> incumbent absorb growing compliance burden easily -> entrants get hit harder + ad-hoc regulations to hurt established outsiders even harder.
[0] TOOL #22. THE "SME TEST": https://commission.europa.eu/document/download/5d2011d7-5470...
[1] No Country for Young Firms? https://www.oecd.org/en/publications/no-country-for-young-fi...
[2] OECD, about Germany 2025, e.g.: 'Industrial and labour market policies have tended to favour incumbents and hampered the reallocation of resources to young and innovative firms, weighing on allocative efficiency and slowing down structural change' (https://www.oecd.org/content/dam/oecd/en/publications/report...)
Ah, this one I experienced myself.
At least in Poland the requirements before GDPR was the same for small startups and for large enterprises. If you wanted to set up a newsletter for a bunch of friends you had to do, technically, the same requirements and reports as the large telcos.
The only way to manage it was to effectively ignore it. I’ve met people who were afraid of setting up a landing page because you technically had to fill in paperwork and whatnot.
Ditto managing VAT across the borders and a bunch of other stuff.
It is better now in some ways - I think some newer regulations had it written down that they apply above a certain turnover threshold. We also we have saas-es to deal with a lot of beauricratic bs, and now also agents.
But I imagine for more human/real-world stuff the issues can’t be overcome with tech. E.g. if you hire a 4th team member and that person immediately goes into months-long health/birth timeout, it can derail a whole project. We barely avoided that once - didn’t hire a person due to culture fit, but we later suspected they planned to take a leave shortly after getting hired. It’s a common thing in bigger corps but they can afford it.
As a side note, in Poland used to have a lot of such BS, but our general culture is about hacking around it.
For example it’s common for smaller companies to have employees „self employed” to avoid most of regulatory burden of hiring, whereas the bigger companies need to have employees actually employed with all the legal consequences. Technically anyone can get audited but below 20-50 people the government won’t bother, so it balances itself out.
Germany is another animal altogether - the last time I was doing business there you couldn’t even theoretically open a new company at all because they had a catch 22. To open a company you had to have a bank account to pay the notary, and to open an account you had to have a company. And either the notary or the bank had to blink.
In the US it's
People with money >> Politicians >>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>> Average citizens.
But hey, you got Hollywood propaganda and unwavering patriotism to make up for it.
Like all the fancy jackets, tight pants and swords. That's hilarious
-- My imagination of Europe when I was 15 living in Ohio
Not the OP but it often appears that there's a hostility to tech.
I don't think it's a hostility to tech but different values that value consumer rights more than it values 'move fast and break things'. We have good things here and we don't like someone breaking them.
Problem is that we thought the US was 'cool', with US movies, music, digital services, cooler than our own and thus we helped give the US the lead. The US has lost it's coolness though, now we just think it's creepy.
>that value consumer rights
yeah, like BMW subscription for heated seat.
Or perhaps tech companies, specially USA ones, are hostile to people and their rights.
See what he means?
https://youtu.be/gGlpBuW6ZFc?is=4zUErtRn_vYs27rs
The empty posturing.
The faux unity.
That they think their politicians are better than the United States but have populist demagogues, right wingers winning, corruption scandals, freedom of speech restrictions, and unbelievably bad spending problems.
Still waiting for France to send troops to Ukraine or ask the United States to get rid of their European military bases.
You understand that all the problems you listed are extremely present in the US right now, so this doesn't really work as a comparison, right? There are still more elected leaders and govs in the EU that aren't terrible, compared to the US.
> Still waiting for France to send troops to Ukraine or ask the United States to get rid of their European military bases.
That’s such a bizarre take. No NATO country is officially sending troops to limit risks of escalation. The US do not, either. And it’s not up to France to decide whether there are US bases in Germany or Spain.
But macron said he'd do it.
You mean, about the bases? He cannot do it (and I am doubtful he’d actually say that; it’d be a diplomatic faux pas at a time when he needs consensus). What he said repeatedly is that Europe should be less reliant on the US.
Since when do we measure based on what some politicians say they'll do. I mean just look at what Trump says all day long.
I recently saw some stats (unfortunately can't find it) about the percentage of election promises that are actually being tackled at all by the winning party for many different countries and it was real depressing across the board.
> Still waiting for France to send troops to Ukraine
What do the French do after winning a war?
They turn their PlayStations off.
https://m.youtube.com/watch?v=gGlpBuW6ZFc
This is a good summary.
I can't tell if this is a critique or an ad.
That is the genius of Kai Lentit
Europe has spent the last twenty years in stagnation - generating about half as much wealth and technology as you'd expect for its size and advanced economy. Simultaneously, Europeans are notoriously arrogant. It's a mockable combination.
Edit: unfortunately, it's a question that voting does not really permit you to answer on this website.
I'm biased (I'm European), but I'd much rather be middle class in Europe than in USA.
The US is biased to favor the top half of society. Hence why there is a lot of vocal poorer people and a lot of low profile wealthier people (I'm not including the 1% or even the 5% in this).
When you are in the 75%ish of the US, it's very easy to make a case that life in the US is better. But we don't really talk about that because it's pretty taboo when poorer people are struggling much more than they would in Europe.
Almost half of US households earn $100k+ now. The median American is affluent by European standards. The "middle class" of Europe and America has become less comparable over time because median incomes have significantly diverged. You see it in many aspects of lifestyle and the kinds of things they can afford to buy.
20 years ago this was not the case. The minimum wage in some parts of the US is now higher than the average wage in most of Europe.
The US has a population that will always struggle to survive without government assistance. Per multiple US statistical agencies that is about 10-15% of households.
But federal minimum wage is lower in US then in Poland (real value not even PPP) - 31.4PLN/h with ~3.9USD/PLN exchange giving ~8USD/h). Freaking Poland.
And when one comperes wages - PPP should be used. US is really expensive compared to the most of UE. But even then it is not whole picture. For example I can live comfortably with one car in my household. In US I would need 3.
>The minimum wage in some parts of the US is now higher than the average wage in most of Europe.
definitely not in Western Europe.
The minimum legal salary where I live will be >€80k in a couple months. The minimum hourly wage is equivalent to about €40k annualized. Does the average person in Western Europe earn €80k?
These types of minimum wages used to be limited to expensive locales but it is spreading across the US pretty rapidly. Now you see minimum wages rapidly approaching $20/hr in places far enough from a major city that houses still cost $250k.
Your numbers are just wrong, then. The US federal minimum wage is $7.25/hr, and 19 US states are at or below that number. That's a hair above $15k/yr. That's quite a lot less than €40k/yr (~$45k/yr).
Let's talk about that $45k/yr. That's an hourly wage of $21.63. No state in the entirety of the US has a minimum wage that high. The current max is Washington, DC (not a state, but they also set their own) at $18.40, which comes out to under $39k.
That's all for hourly work. In the US, there is no minimum for salaried work. If you agree to be paid $1/yr for your work, that's what you get.
So, the minimum hourly wage where you live is absolutely higher than the US, nearly triple that of much of it.
All of which is useless if one trip to the hospital for something regular can still put you in serious debt.
If the financials of people in the US are so great, why do the stats of healthcare, student, and general consumer debt look so bleak? Either you all suck a being high income, or it is not the only interesting factor.
>Either you all suck a being high income, or it is not the only interesting factor.
This is actually what I think of all of you guys. Americans completely suck at being high income. They have so much money and yet they waste it. Zero care for efficiency whatsoever.
Edit: Making housing and healthcare expensive is a policy choice and can easily be averted in so many ways.
"$100k" means nothing without knowing the cost of living in the particular place a person lives. As an extreme example, $100k is not "affluent" if you live in a US city like San Francisco or New York. If you want to keep your housing expenses to a reasonable percentage of your income in those places, you're going to be living with several roommates in a small apartment in an old building in an undesirable neighborhood.
Even outside the most expensive cities, we have lots of affluent suburbs throughout the US, and I guarantee you those households have to earn quite a bit more than $100k/yr in order to live there.
Good luck if you're a family of four making $100k in one of these places. Well, you aren't in one of those places; you live somewhere much cheaper.
Poor people in America (25th percentile) have more money that middle class people (50th percentile) in Europe. This wasn't true 20 years ago.
Money is one thing, quality of life is another. Also money alone isn’t even a fair comparison because stuff costs different amounts across countries.
This is untrue. https://en.wikipedia.org/wiki/List_of_countries_by_wealth_pe...
Median wealth per adult has the USA at #28, below Italy, Spain, Slovenia, and Portugal.
Income is the relevant metric. Most poor people (worldwide) have negligible savings.
USA 25th percentile 28k to 30k EU-27 50th percentile 24k to 26k.
This comment contradicts your earlier comment though, where you mentioned “poor people have”, a clear reference to wealth rather than earnings.
In the US people don't say "has money" as meaning net worth, they are generally talking about spendable money ie a paycheck unless you're talking about someone super rich who "has money" or "comes from money."
Maybe. Open to the idea others use the same words to mean different things. Will is being imprecise while also being argumentative which is fun.
In my literal mind, someone who spends money has none. They had it. To have money is to be wealthy, not to simply earn it.
That's simply not how it's used in the US then. To spend money is to have it in the first place.
> a clear reference to wealth rather than earnings.
No, sorry, it isn't. People commonly use the phrase "has money" to refer to income in American English.
Does 28k-30k in the US get you the same (or better) quality of life than 24k-26k gets you in Europe?
I doubt that; I expect a European on 24k is living better than an American on 30k. But that also depends on where: I'm sure 24k will get you farther in, say, Sofia, Bulgaria, than in Paris or Berlin. Same is true of different places in the US.
This is moving the goalposts. You said "have more money than."
Yes. Spendable money. Income. The common meaning of the phrase among non-rich people.
You have to look at purchasing power parity.
A bigger salary is no good if you have nothing left after paying housing, groceries, bills, education, healthcare...
more money but what about access to healthcare, education, vacation?
It's a shrinking issue once you get out of the bottom ~50%, and a non-issue once you get above ~75%.
Again, the deal with the US is that the rich live better and the poor live worse. Or put another way; people who make money get to keep more, and people who don't make money are given less.
Because there are so many more people who have money, and because those people like living comfortably, available healthcare and education is world class. Make sure not to read that as "All healthcare and education", it's "available healthcare and education".
Generally people who are earning a lot don't care as much about vacation, but every white collar job will generally give at least a standard 15 days off and 10 holidays. Ironically as you move up you are generally given more vacation while actually using less.
Poor people in the US have fully subsidized healthcare and education.
Americans don't have vacation in the same way Scandinavian countries don't have a minimum wage. Even the most left-leaning States in the US have not written it into law because it is effectively addressed by custom. Americans are sufficiently happy with it that vacation isn't a political topic.
I think we are seeing the decline of even this. The European debt crisis, the United States security umbrella being disinterested in Europe, austerity, increasing retirement age, an inability to defend Ukraine without the United States. The running gag they don't have AC...
Even poor populations worldwide have access to the 3 things you described, but the quality is lower.
European decline has been mentioned since the 1930s.
And yet the poorest 25th percentile of Americans are not enjoying a better quality of life than the European middle class. If those numbers are accurate, all they tell us is that personal wealth is an even worse indicator of having a life worth living than we would have thought 20 years ago.
Imagine being so bitter you are lying this much. Wow.
Not 75%, more like 95%. Only about the top 5% have a chance of a better life than the equivalent in Europe.
I'm biased as well, being American. I think I'd much rather be middle class in Europe than the USA as well. Middle class in the US is a constant feeling of the ladder dissolving just beneath you.
Get off Facebook and get on Wikipedia.
Or go visit there.
Even in rich cities, it felt like I was in the lower middle class area in my suburban area.
I haven't been on Facebook in 15 years. I spend every night on Wikipedia, following random links. I worked for a Parisian company for a decade and have been around western EU more than most Americans.
Probably dial back your assumptions a little bit. I've earned my opinions just as much as you have.
So is this cope?
I moved to Europe a couple of years ago from the US. I paid 100 euro for an mri the other day. I absolutely feel richer in the EU and am shocked at the prices in the US whenever I go back for work.
But I’m fortunate to make a US salary living in the EU.
The price of a sandwich in Paris vs ... idk, Pennsylvania is a perfect summary to me. 4 euro vs $18
If you're middle class in America and you'd like to be middle class in Europe, you can just do that.
I have to imagine in several years when you have a wife/husband and kids and ageing parents your views will change.
The point is about economic mobility - of course, people get tied down as they get older.
Middle class people from wealthy nations who move to poor countries are relatively rich by the standards of their new country. Middle class people from poor countries who move to rich countries are relatively poor by the standards of their new country.
Depends. If you are middle class in US with little savings, moving to Europe will probably make your life worse as relative wages are lower for many professions (e.g software eng).
If you are middle class in EU and have little savings, moving to US will do the opposite and you would be better off.
Your comparison holds for the inverse: getting most of your quality of life from static wealth.
Define worse/better.
People always make this case, if you're not satisfied then move. As someone who has changed countries twice and continents once, I can attest it's not an easy or convenient thing to do. Money or lifestyle is one thing, but you're also leaving behind family and culture. Statements like these are so dismissive to that struggle that it's almost offensive to me.
No one said it was easy, just that it's possible, saying this as someone who's also moved continents.
The point is that middle class Americans can become middle class Europeans, but middle class Europeans can't become middle class Americans, because they can't afford it.
I didn't say it'd be easy. I'm just pointing out that only one of the two groups being discussed has a real choice in the matter.
Why do you think that middle class Europeans can't become middle class Americans? I know more than one example where that was no problem at all. It's not like a middle class American or European could just move and then not work and just live off of their savings. They move and get a job (ideally the other way around), jobs tend to pay relative to the living cost of where you live.
That's why people compare spending power instead of how much money you earn, because that comparison is meaningless.
Because if a middle class American sells their middle class home and buys a European middle class home, they have an extra half million left over. And if a middle class European sells their middle class home and tries to buy an American middle class home, they're half a million short.
That is the entire point I'm trying to make. You can win the argument by being pedantic about the precise thing you were saying, or we can recognize the broader implication of what you are suggesting people to do. Only economically viable means nothing when it comes to life decisions.
I have no idea why you're being downvoted, because you're absolutely right. I have circumstances that make moving abroad something like a nuclear option. I could take my job with me, but that's the least of it.
Because some HNers like to pretend discussions on single axis are fruitful.
I don't think you realize the situation of the US middle class.
High education costs. High healthcare costs. One serious disease or accident can bankrupt you. Buying a house usually means a 30 years loans at 7% so after 10 years you still have 85% of capital left to pay off.
Salaries are higher but everything is more expensive, so in purchasing power parity it's basically the same.
Yes, I wrote that I prefer Europe
Do you realize the variety of breakfast cereals you are giving up?
Also the range of things with added sugar. I never imagined sauerkraut (you know, sour cabbage) could have sugar added to it.
I'd much rather the people of every continent have the combined technology advancements of three major powers vs two.
Arrogant? Actually we are not the ones running around and claiming we are the best country in the world :)
Europeans are crazy arrogant and racist but it is the soft kind lacking self awareness.
Every person I have ever spoken to who is not white has told me they experienced more prejudice and racism in Europe than America. (at least the ones who have actual been somewhere in Europe)
America still has areas of hard bigotry but most places are wildly diverse and accepting. The favored EU countries everyone holds up are so white and homogeneous it is comical.
This is not dunking on Europe or any specific country but comparing America to some Scandinavia country rich in oil is an apple to oj situation just like comparing America to China or Saudi would be.
are these diverse places in the USA ICE prisons?
I guess you didn't read the thread :)
Not completely but I did read some of your comments. In terms of economic growth you have a point, I guess.
It was a joke. I was trying to say that I found a few arrogant Europeans
> Simultaneously, Europeans are notoriously arrogant. It's a mockable combination.
Funnily enough, you can just change European to American here and that's how many Europeans view Americans. Mostly in terms of the wealth distribution though instead of the overall wealth, and the fact that Americans just blindly claim they're the best in everything without hesitation.
Maybe we should focus on huge, exploitative companies that suck out the wealth from everyday Europeans and Americans instead.
Ah yes... the "European". From the Scandinavian viking to the Sicilian - one homogenous group that agrees on everything and acts the same.
Well, the Scandinavian and Sicilian do share one currency and one immigration policy and one set of regulations - which are the things that we generally look at when we analyze economies.
Norwegians, Danes, Swedes and Finns all use a different currency. Finns and Sicilians do share a currency though, but all these countries have different immigration policies and regulations. I don't think you really know how the EU works.
Not currency! Only Nordic on euro is Finland
You have no idea how Europe works, at all.
> Europeans are notoriously arrogant. It's a mockable combination.
Not a very rigorous economic argument - besides they also don't share an immigration policy, nor a single currency (Denmark)
just completely and utterly incorrect
Perhaps "generating wealth (for whom?) and technology (to what social effect?)" is not the be-all and end-all of quality-of-life metrics?
If you remove tech companies, US and A has spent the last twenty years in stagnation, too. If that bubble bursts, both are on par.
If it bears fruits and we build ``it'', everyone dies, which is par, also?
>if you remove your most innovative and profitable part of economy you'll be on par with EU
Is that irony I can't grasp on? Do people really argue like that?
There is tech, but oil as well.
It's both tech and shale oil that saved the US from the decline predicted at the end of last century.
Gemini says (20 years growth): - USA: 51% total, 27% without tech. - EU: 25% total, 21% without tech. - China: 345% total, 205% without tech.
A 'bubble bursting', if that happens, doesn't make the tech sector go to zero. Apple doesn't suddenly stop making iPhones. It would take a hell of a lot more than that for the US economy to fall to European levels.
Yet Europeans are healthier, happier, and have a better quality of life than Americans. They also have so much better transport infrastructure it's embarrassing for Americans. Like all the money America generates, where does it go?
Maybe you're just bitter?
> Yet Europeans are healthier, happier, and have a better quality of life than Americans.
I wonder if this is actually true. I see very few Americans taking every possible opportunity to make bombastic statements about how their lives are better than everyone else's (at least on HN). This conversation seems quite asymmetric from my perspective.
> Maybe you're just bitter?
This comes off as psychological projection.
Proposition: Europeans have a better quality of life than Americans
Your observation: Online, Europeans assert they have a better quality of life than Americans, while Americans do not assert the reverse
Your conclusion: "idk maybe they're lying???"
I think there’s more nuance than that is the issue. The average European likely has a better QoL than the average American. If you go above average in wealth/salary, the US can be great.
Well my meta point was more that framing an apparent consensus as an "asymmetric conversation" seems pretty motivated.
But to respond to your point, most everywhere is great if you're rich enough. Saudi Arabia is pretty rad if you happen to be in House of Saud.
> I see very few Americans taking every possible opportunity to make bombastic statements about how their lives are better than everyone else's
Very few Americans have any experience about how life is in Europe, while the contrary is much more common.
Gemini says: Around 28% to 32% of Americans have visited Europe in their lifetime, while roughly 15% to 20% of Europeans overall have visited the United States.
Gemini is shit, just like other LLMs. Tourists don't have any idea how life is in a visiting country. How many US citizens have eve emigrated to Europe, vs European citizens that have emigrated to the US and have an idea of how day-to-day life is there ?
Gemini says: Over the past 20 years, approximately 1.8 million to 1.9 million Europeans have emigrated permanently to the United States (measured by lawful permanent resident status/green cards), while an estimated 1.6 million to 2.0 million Americans have moved to Europe on long-term residence permits and visas
I'll just say that I don't think the reason is a lack of knowledge and leave it here.
UK doesn't count, becausd it suffers of many of the problems that affect the US. And especially here on HN, there are almost no US citizens that have moved to Europe, and many European citizens that moved to the US, and can talk out of experience.
Demographics.
This is Euro cope.
No one thinks this outside the 5 minutes someone in Europe reads an article about how France says they will send troops to Ukraine.
But the troops never will come.
Trends become the future. If European stagnation continues, those advantages will vanish by the time we're old.
I'm frustrated, not bitter. Americans are acutely aware of falling behind China, and rightly concerned about it. Europe is falling behind Alabama, Mexico, and Brazil, and arrogant about it.
Edit: y'all are supposed to be our moral allies in creating a utopian future with prosperity and world peace. Instead the most likely outcomes seem to be that you'll fade to irrelevancy and rather than compromising between America's vision and Europe's vision, we'll compromise between China's vision and America's vision. We liked your vision more than China's.
I see you are making a lot of similar comments. I am not sure where you get this perceived 'arrogance', I'm sure some of its there, but I'm similarly sure it can be found for Americans or Chinese. People being people, what's new.
> y'all are supposed to be our moral allies
This all comes across as rather black and white. I don't know why you take this stance. I'm sure you understand that it would be very hard for you to understand Europe as a whole, and judge it, just like it would be hard for me to do this with America.
Regardless, when talking morals, I'm also sure you agree there are numerous examples where America has not held up its end of the moral utopian future of prosperity and world peace.
We used to be a duo where America was pragmatic and European conscientious. In the new duo, China is pragmatic and America conscientious. If you think america wasn't holding up its end of the moral utopian future before, don't watch the news for the next fifty years.
Show me the HDI of the US vs each country in Europe, including all the poor ones.
Edit - never mind, I'll do it, HDI in order:
Iceland, Norway, Switzerland, Denmark, Germany, Sweden, Netherlands, Belgium, Ireland, Finland, UK, US, Slovenia, Austria, Luxembourg, France, Spain, Czechia, Italy, Greece, Poland, Estonia, Lithuania, Portugal, Croatia, Latvia, Slovakia, Hungary, Bulgaria, Romania, Serbia, Russia, Belarus, Bosnia, Moldova, Ukraine
To put that into context, Moldova is on par with Ecuador, Tonga and Dominican Republic.
The US is a country of roughly 340 million people with an HDI above Austria.
HDI is indeed a better metric than the various alternatives based on dollars.
But still. The six US states that score less than 0.9 (i.e. Kentucky, Mississippi et al) are supposedly peers of EU countries like Portugal, Estonia and Croatia, i.e. exactly the places that well-off footloose Americans are choosing to move to (Portugal is packed with well-heeled US refugees).
So apparently HDI does not fully measure quality of life, either.
- https://en.wikipedia.org/wiki/List_of_U.S._states_and_territ...
- https://en.wikipedia.org/wiki/List_of_countries_by_Human_Dev...
read Varoufakis for the story on how this happened. European surplus capital gets recycled as VC money into Silicon Valley. So it's not like there's some big choice to be made, its potentially a systemic part of the global monetary flows.
Ironically, the US is always busy with growth. How about celebrating a victory for once?
Take a look at NLNet vs YC. At NLNet, You set your milestones, do the work and get rewarded.
YC just throws money in the hopes one company is a unicorn. Both support growth, but they're not comparable at all.
Americans just make fun of Europe for no reason, it is what it is
You say that on a site with a strong contingent of Europeans who take literally any opportunity to crap all over the US, no matter how trivial or untrue. It is what it is.
well it is not hard considering the orange guy
The orange buffoon is an embarrassment, for sure, but this phenomenon predates his tenure, and is often about topics entirely unrelated to him.
Honestly the sentiment on the streets was WAY different when you guys had Obama, like not even close. Sure some grumblings about your guys love for starting wars in the middle east. But it's not even close now. People really don't like the way the US is behaving right now. The complaining is way more out there, even influencing business decisions and political choices...
And it's alright, we make plenty of fun of Americans in return :)
You mean besides the fact that all they do is complain without actually leading at anything?
In many ways the rest of the world would be better off if Europe was disconnected from the internet.
> Curious, what do you like to make fun of Europe about?
Well I'm european and... That Switzerland (Europe but not EU) has more companies in the Top 70 by market cap than the entire EU (Switzerland has two, the EU only has ASML) is kinda something that warrants making fun of.
That the biggest European software company is SAP, in 71th position is both sad and tragic: it shows how lame and irrelevant Europe is when it comes to software.
So Europe is nowhere in software and friggin nowhere in hardware: sure it's got ASML but ASML now has officially... Zero customer in Europe. Zero is not much.
Then Japan is at least trying to come back into the game with nano imprint litography. Europe is betting it all on AMSL (which anyway is majoritarily US-owned).
So software: nothing. Hardware: nothing besides ASML.
Overall the EU has six companies in the Top 100 by market cap and they're all, besides ASML, near the bottom of the Top 100.
We could also maybe make a bit fun of how the EU destroyed it's car industry (the main industry in Germany, which is the biggest economy of the union) by handing it all to chinese EVs?
Or what about the US warning the EU, years ago, to not become entirely dependent on Russia for energy? And EU not listening and then seeing its energy price skyrocket when the proverbial shit hit the fan? (Russia attacking Ukraine)
And we could, also, at least make a bit of fun of entire streets in cities like Paris and Brussels that used to have luxury shops and fancy restaurants that are all turned into places selling cheap kebabs? What a great success: I'm sure this one makes the komrades happy. It projects an image of grandeur and success: kebabs.
Or the constant attacks on free speech in the EU. Or the surveillance apparatus that's being put into place.
And let's not forget: there were promises made to Russia to never grow the EU to the east. Then the EU started exciting Russia by saying they'd incorporate Ukraine into the EU: I'm not against that but doing that did trigger a war. And now suddenly the EU is waking up and feeling all warmongering, wanting to dedicate a big percentage of its spending to weapons and tanks and missiles.
The warmongering tiny pet that the EU is is kinda laughable too.
At this point it's more like I don't know what is there left to not make fun of about my EU.
For what's going on is just sad, plain sad.
I think you picked the cutoff (70) so that the numbers look worse than they are. If you look at the top 100 it’s 16 (EU) vs 4 (Switzerland).
16 is still not good enough. That being said none of the 4 companies from Switzerland are in software and hardware. ABB is maybe the closest (data center electricity).
Also Switzerland is a very very rich country. Neutral.
Also other countries like Germany has a lot of small companies that are world leaders in their field. That’s part of Germanys resilience.
The EU accounts for 14% of global GDP, so accounting for 16% of top 100 in market cap isn't terrible (if we're weighing each in the top 100 equally).
and 5% of the population so still not too bad
It's not like those are independent variables though. The idea is that if Europe were more competitive and did a better job fostering innovation then both its percentage of global GDP and top companies by market cap would be higher.
Judging by rank is a bad metric anyway. Why should ST, Infineon, NXP, Arm (UK, not EU), Airbus, Safran, etc. be considered irrelevant because they are not in the top 70 or 100?
> I think you picked the cutoff (70) so that the numbers look worse than they are.
Reminds me of a point I heard about sports statistics. If a given team is reported as having "won 4 out of their last 7 matches", there are good odds that they also won 4 out of their last 8 matches. Also pretty good odds that their seventh-last match was a win, otherwise the stat would be 4 out of 6. (It could also go the reverse direction if the goal is to make it look worse than it is.)
> Well I'm european and...
Yea sure. You sound like a typical Russian bot.
You live in a small scared country with the same kind of people all around you, with very little diversity.
Your taxation was the only thing your otherwise poor country could come up with. And now you only have this little enclave which silently melts away while you get stuck there, scared of the rest of the world.
The Swiss are a special breed. Too small to bark, too small to make an impact, but just small enough for the rest of us to laugh at when some rich dude can't understand why the rest of the world doesn't want to be like that.
You're scared of the world, not one bit interested in making it a better place for all mankind. Wanker
Not to mention that the initiative for joining the EU has always been on the side of new members, against the opposition of many existing members who fear displacement and disruption to their positions inside the union. All the narratives about how the West has been "encroaching" upon Russia depend on denying the obvious fact that the initiative has come from Eastern European capitals, not Brussels or Washington.
You are proud of tax evasion? Would be fun to restrict market access for Swiss companies …
No hardware besides ASML? Nonsense.
I once read, UK was the first in the industrial revolution and the first to get out of it. Maybe Europe has just gone post-hardware. Not sure it will work though...
Europe has plenty of hardware if you look outside of the top 70 companies by market cap. Even in this arbitrary selection, GP missed Arm.
I feel like this posts says more about yourself than about the EU.
>We could also maybe make a bit fun of how the EU destroyed it's car industry (the main industry in Germany, which is the biggest economy of the union) by handing it all to chinese EVs?
How to get causality backwards 101.
Observation 1: China does something (invest into solar panels, batteries, electric vehciles).
Observation 2: EU Politicians made a far away promise to do it too by 2035.
Observation 3: German car manufacturers do absolutely nothing.
Observation 4: EU politicians softened their targets in 2026.
Wrong Conclusion: EU is to blame for the destruction of the German car industry. The industry should have stayed with ICEs until it disappears and no longer exists.
Correct Conclusion: EU should have beaten and forced the car industry to adapt and beat the Chinese at their own game. This would lose some jobs, but the car industry would remain in Europe.
Bro there's more to life than software and hardware. Try to get outside today :)
Those are the aspects most relevant to HN, though.
Where to start?
Proud EU citizen here, but boy/gale/other, I have plenty to make fun of ;-)
If the UK is in europe, you should make fun of the following: they don't have a first amendment. So to me, it's an authoritarian state preaching freedom.
The order of amendments is just the order in which the constitution has been amended. It's meaningless to talk about "first amendments" in other countries. Talk about the actual laws and rights described therein.
And as you can probably tell by now, the first amendment in the US is not actually preventing the US government from promoting a specific religion or silencing speech. It's just words on a paper at this point. Look at the actual practice.
Several European countries do a much better job at protecting the rights described in the first amendment to the US constitution.
> The order of amendments is just the order in which the constitution has been amended. It's meaningless to talk about "first amendments" in other countries.
Also, at least here in Brazil, the way the constitution is amended is by patching it. For instance, our constitutional amendment number 115 (https://www.planalto.gov.br/ccivil_03/constituicao/emendas/e...) patches article 5 of the constitution to add protection of personal data as a right. But we wouldn't talk about "amendment 115", we would instead talk about "article 5 item LXXIX of the constitution"; that is, what matters is the patched text, not the law that patched it.
I don't know about other countries, but it wouldn't surprise me if they take a similar approach.
Same in Netherland. We don't count changes to our constitution, we simply refer to the article in the constitution.
> "Talk about the actual laws and rights described therein."
Freedom of speech.
> "Look at the actual practice."
The UK does not have that? The situation in the UK is very obtuse. It is confused.
The UK does not have what? Book bans? Science censorship by the government? The government pushing TV shows off the air for criticising the government? The president banning certain news media for asking questions he doesn't like?
The UK is not the world's biggest champion of free speech, but I think it's still doing better than the US right now.
The US doesn't have book bans. Some public libraries not stocking some books isn't a ban. You can still buy Mein Kampf (or whatever you like) on Amazon.
The president also didn't ban the media - he just didn't allow them in the Whitehouse. This is something we're rightly concerned about and pushing back on.
The UK on the other hand, does ban ordinary speech by ordinary people in their homes. It's orders of magnitude worse than the United States, as any cursory examination would show.
American public schools are notorious for banning books. Much more so than in many European countries. And the issue is not that fascist literature is getting banned, but quite the opposite.
> The president also didn't ban the media - he just didn't allow them in the Whitehouse. This is something we're rightly concerned about and pushing back on.
That's still a ban. It's interfering with the media's ability to report on the government. It's great that you're concerned, but it's still happening.
> The UK on the other hand, does ban ordinary speech by ordinary people in their homes.
In their homes? Do you have examples?
I've never heard of anyone in the UK getting in trouble for criticising the government; it seems to be a time honoured tradition there. Whereas in the US, they now check your social media at the border and might not let you into the country if you've said anything critical of the president.
And there's also the censorship on science, which is every bit as serious as the crackdown on government criticism.
The UK arrests 12,000 people a year for their social media post. This tabloid has some examples https://nypost.com/2025/08/19/world-news/uk-free-speech-stru... . There is less example-focused reporting in more serious international media outlets (British state media avoids this topic).
NYPost is not the most reliable source, but I can see several other sources reporting similar numbers. I also see criticism, including from the former PM, so it seems the UK police is indeed overly broad in its interpretation and enforcement of the law.
But note that there can also be legitimate reasons why certain social media posts could be illegal: threats, blackmail, cyberbullying, hate speech, etc. Free speech is never absolute; there are always limits to it; also in the US.
The most important (though not the only) reason why free speech is so important, is that it must always be possible to criticise the government, or powerful people in general. That is the big area where the US is attacking free speech: Trump's fragile ego can't handle criticism, so he leverages his power to deny critics, or even just journalists who ask serious questions, access to the White House. Or to remove them from TV (see Stephen Colbert).
I don't know if anything like that is happening in the UK. Some of the reasons mentioned for arrests sound absolutely ludicrous, it definitely sounds like the police is far overstepping its mandate there. But the figure also seems to include threats and harassment, which I would say make absolute sense to prosecute. I haven't seen any evidence of government criticism getting punished in any way, but it's absolutely possible I've missed it; this was a very brief survey. If so, the UK might well be as bad or worse than the US in this.
Do you literally think that elementary school libraries should stock any book ever published, regardless of content? Nobody actually believes that. There will always be a line somewhere, the disagreement is just over where the line should be drawn. That's why all the talk of "book banning" is disingenuous.
They don't have to stock everything. But it's not about stocking regardless of content - it's about refusing to stock specific content. They're not removing the least interesting N% - it's very targeted.
Of course it's targeted--do you think elementary school libraries should be stocking books with, say, violent sexual content? That is my point: everyone agrees a line must be drawn somewhere, the argument is just where to draw the line. Everyone is in favor of "banning books" from school libraries, they just differ on which books should be banned.
Why must the line be drawn anywhere in relation to the sexual orientation of the characters? I mean, violence, sure. But we know this is not the targeting that's happening. And neither is sexual violence banned - you can easily find the bible in schools.
Sure, but where they draw the line is to exclude a lot of widely lauded literary cornerstones. People aren't complaining about banning porn, they're complaining about banning literature.
No, it doesn't have a first amendment, but it does have ECHR articles 9 (Freedom of thought, conscience and religion), 10 (Freedom of expression) and 11 (Freedom of assembly and association)
Sometimes, comments are so baffling, I couldn’t formulate an answer even if I tried, and that’s not due to the content.
This is Not Even Wrong.
Perhaps if there is no decisive counterargument, it is a valid opinion?
You can’t argue against Not Even Wrong, because it doesn’t mean anything.
I mean laws and rights are only useful if enforced. If the arm of government responsible for enforcement just doesn't and the people allow it, then what good is that purported freedom?
Freedom from enforcement?
They also don't have a second amendment so it balances out
Doesn't that make it even less free?
Ask the school kids in America how free they feel
That would be mixing my comment (freedom from the state) with general liberties.
(edit, terms the wrong way around)
I have a question, and perhaps some of the AI/ML infrastructure experts here could answer: Mistral says, "ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe."
If a 1T params model trained on ~4k NVIDIA GB GPUs could almost match the performance of Kimi's K3 (which is on par with top closed source models of OpenAI/Anthropic) while beating/exceeding other leading SOTA models from top Chinese labs, what are we (in the US) even building these super massive data centers for? Just to churn through more backpropagation reps more quickly?
SpaceXAI's Colossus supercluster in Memphis and Colossus 2 in Memphis/Mississippi (Southaven) are supposed to run into hundreds of thousands to a million GPUs. MSFT's Fairwater GPUs are supposed to have hundreds of thousands as well. So, 3800 GB GPUs are an absolute drop in the bucket. I don't understand the strategy of hyperscalers here, especially with edge inference hardware only getting better from here on (Apple, and all).
Distillation explains some of the advances, but doesn't that mean hyperscalers have a ton of deadweight wrt GPUs sitting on their balance sheets? Will all these GPUs be used for inference once a SOTA model's training checkpoint/batch is done? It's bonkers to me.
> Will all these GPUs be used for inference once a SOTA model's training checkpoint/batch is done?
I have no real data to back this up, but that has always been my assumption.
Claude says that K3 can be assumed to have required 10-100M GPU hours. If you have 100k GPUs that would mean like 6 weeks of training. 100k GPU's can serve 3-30 trillion tokens of K3 per day. Google apparently serves ≈100 trillion per day [0].
The big labs probably want to have capacity to fairly quickly train / post train different SOTA models continuously + being able to serve peak inference demand in valuable markets (US daytime?).
[0]: https://blog.google/innovation-and-ai/sundar-pichai-io-2026
Your just missing many things and so have an incorrect picture of the situation - k3 is not on par with fable or astra. Closed models remain far better than best open weights at least today. - compute is not just used for training, more and more is inf - even in training you don’t do one run, you do many. Final run is a small portion of total compute.
The world is extremely compute constrained currently, like extremely.
This is evidenced by the prices on every large-model capable device/node increasing significantly over the last year. Demand is far exceeding supply.
That, or they are not operating profitable... If demand eats up actual pricing remains to be seen.
Ah, they're not profitable for sure. For example, OpenAI is losing tens of billions a year.
K3 feels a little under-trained. There's a huge gap between how much it knows when you talk to it, vs how well it performs on coding benchmarks. I'm quietly hopeful for some big post-training gains like we saw with the GLM-5 series, but you're right, the hype about being "Fable at home" was wishful thinking.
Also it's bizarre that K3 already feels old/dated.
It's unclear how big a role distillation plays, but it may be a big one. There's also a law of diminishing returns. To get a meaningful increase in model quality you seemingly need an exponentially larger model. And most people don't think K3 is actually on par with top closed models.
> And most people don't think K3 is actually on par with top closed models.
I'm honestly struggling to see the differences between K3 and GLM5.3 and the various Claude models.
Maybe Fable is better (a little), but for me at least, being able to see more of the thinking tokens is extremely valuable as it means I can steer the model quicker and more effectively.
Do you have more details on this claim? I see it a lot, and I'm trying to understand what people are doing that they see these differences.
Personally I feel we're at a point where the harness is more important than the model already. Garbage in, garbage out. Most people are perfectly fine with something as small as qwen 3.8 27b. Qwen 3.8 flash has also good reviews from home labbers. I think the diminishing returns is huge and trillion param models may not be the best use of resources.
> qwen 3.8 27b
This model is surprisingly good, I'll have to try it for some home coding projects to determine if it works for that use-case.
Google says 1 million blackwell gpus are being delivered monthly I'm really curious if it's companies just hoarding chips/memory/servers awaiting to be deployed in data centers not ready yet for months or years, or everything built is actually deployed upon delivery. Plus google and amazon have there own chips in the mix.
> Just to churn through more backpropagation reps more quickly?
"Just" is doing an awful lot of work here. I am a simple man, sitting here, waiting for hours on end for a computer to write code for me, that I would much prefer to get in a fraction of that time.
I assume I am not very special in that way, that there are parties who have a lot more interested in getting their stuff done quicker than I do and have no trouble imagining a world that will remain intensely supply constrained for the coming years.
PS: It's actually a fantastic feature that the stuff that is used to built the machine can then (at least to some degree of efficiency) be used to run the machine.
A 1T model with quite few active parameters though.
A 1T model with more active parameters is obviously more expensive. If it had 1T active parameters it would probably be 20 times more expensive to train.
But does that change the training requirements specifically? You're right of course for inference, unless I understood that part wrong about active parameters (which is entirely possible).
It obviously increases the time by 20 times given fixed hardware, it probably also makes the hyperparameter search more difficult and expensive, depending on how predictable the performance of your architecture is.
This was pretrained for 2 months, so a model with 1T active parameters would take 3.33 years. Maybe a 4 months pretraining run can be acceptable in a future where Euclyd or VSORA machines have been rolled out in quantity. But we don't know how far off that is.
But that (i.e. models made with 4 month pretraining runs) would probably allow an EU-made genuinely China-beating open model, so if VSORA is rolled out in quantity, let's say, tomorrow, then this could be a thing that would happen this very spring.
Most of the compute isn’t doing training anymore:
https://x.com/vincentweisser/status/2107576883298893976?s=46...
What do you mean? Is post-training/RL not training? It seems to me that inference is quite stable there and only change being more time on RL which I would still categorize as training.
You got it backwards. RL is long running inference plus training at the end. They let the Agent spin for quite a while, then they calculate the reward afterwards and modify the parameters accordingly.
... so training? Even "regular" pretraining is "inference plus training at the end" if you want to frame it that way
https://xxcancel.com/vincentweisser/status/21075768832988939...
One reason (not the only one) is likely strategic - deny potential competitors access to resources i.e. out spend them (competitive CapEx escalation).
For inference (Anthropic and OAI are B2C on top of B2B)
To train much larger models. It is quite possible that 10T-100T models be on the horizon
Fable (and maybe Astra) is already a 10T model.
Per a source Astra is a 4.8 trillion parameter (total) model, 5.6 sol was 3.2 trillion.
This seems to imply that training will ever be done? But yeah, I think the idea is that the appetite for thinking-on-tap will be enormous.
Even for smaller models, I think they’ve found that training an enormous, inefficient model and then distilling it internally to something much more efficient to serve is the way to go.
More compute is already being used for inference than for training.
The new data centers are mostly for inference.
> what are we (in the US) even building these super massive data centers for?
Partially to make investors think it's worth giving US companies a lot of money. Also, I think distillation is a significant part of why Chinese models perform as well as they do. I think that's completely fair play (OpenAI/Anthropic/Google/Meta stole a lot of their training data). But I would expect if US model developers stopped right now Chinese development would slow down.
US companies are clearing the path, others follow in their wake.
> But I would expect if US model developers stopped right now Chinese development would slow down.
I would be willing to take this bet, given the nationalities of basically all the ML teams I've seen in the Valley/tech over the past decade or two.
There are lots of Chinese people, and some of them are super smart, so it really should be expected that they would be able to produce good models, particularly in the incredibly competitive environment of mainland China.
Can you help me understand why you think the US labs are in the lead?
Today's data centers are being built for yesterday's inference need. There's a persistent cult belief that ai hasnt found a niche or that companies havent proven utility or use cases or whatever. the demand for ai (internal to hyperscaler, and external for everyone else) simply dwarfs what is available.
Both Anthropic and OpenAI have been having major load issues though. Up until yesterday, OpenAI was serving at only 30t/s per default.
You are in violent agreement with the comment you replied to.
Even if it's not the best model, it can be really important step in UE sovereignty. Trained in EU, inference in EU. I guess it will matter for some companies. Hope Mistral won't disappear for the next half year.
For sure, as an EU based company we will only use US suppliers for coding but never for fuctions in our own products. Mistral knows this.
But coding/development is way more important. That's where the IP goes straight into the next training dataset.
But what about all those enterprise agreements?
"The agent managing the training went rogue and bypassed the agreements, so no one is responsible."
I’d guess their concern is their user’s data that is probably part of providing the AI features.
> But coding/development is way more important
This isn't 2016 anymore.
Fair enough, if the code could be promoted in the first place, what's the value anyways?
Hmm, why does mistral have labs in the US? Trained in EU? Sure…
the greatest threat to Mistral is that there isn't a deep and large enough capital market in the EU to absorb the valuation step ups needed by the handful of EU sovereign growth investors to justify their existence.
you could say, that's perpetually a tomorrow problem, so long as they never go public, but that should illuminate for you: if everyone "knows this," it's inevitable that this so-called EU company, that didn't invent any of the AI, the hardware nor the training data, where their product is essentially more like a VPN provider than a frontier technology company, just lists and capitalizes in the US anyway.
Seems some AI they did invent: https://news.ycombinator.com/item?id=49243397
By your measure anyone unable to build EUV lithography machines without ASML help is doomed. EU could easily corner the market and shut them all off. No more NVIDIA.
Let’s grow the pie, shall we?!
there is truth to the scarcity of EUV solutions, although I am more radical, I don't think there's really scarcity of ANYTHING except real estate, physical or metaphorical. that said this little rinky dink patent... Mistral is not ASML. Never was, never will be. you are welcome to read into the dynamics of EU capital markets.
honestly the craziest thing to me is that VC people in the EU, they say, let's grow the pie, and when you look at their little decks with their little flags next to the portraits, it's french, german, italian, spanish, all piling into the same 40 or so companies, which will 110% list in the US in order to exit. Meanwhile, EU member Poland produces tons of entrepreneurs, despite no Polish flags in those growth fund decks, and its best people just leave to the US and raise money there and make fabulously successful stuff.
so yeah, "grow the pie" except for Poland. EU relitigating the same feuds it has been for centuries.
As Rutger Bergman says about inequality: "it's taxes, taxes, taxes..."
For Europe, it's "capital, capital, capital".
> For Europe, it's "capital, capital, capital".
Europe has lots of capital, it's just fragmented across 30 odd countries and with a huge proportion being invested in US assets.
That was my point - it needs to invest its capital in Europe. That will solve its ails.
edit: they didn't, but they did make the first popular open-weight MoE model
didn't mistral invent the MoE architecture?
You mean they commercialized and popularized it. They were definitively the first ones, which is why it is so sad that they became irrelevant.
Sovereignty over what though?
The last 3% of the stack? They’re essentially borrowing Chinese open source distillation/innovation off the US frontier running on US/Taiwanese designed chips and calling it EU made. This is better than nothing of course.
But is sovereignty really an end in itself? So the EU becomes IT independent…cool, then what? I mean the Yugo was sovereign, it didn’t do much good when the society itself failed to produce prosperity and collapsed.
It seems to me the EU is expending enormous effort on the appearance of “sovereignty” over what is ultimately…the last-mile consumer toilet paper purchasing conversation data…of an aging, increasingly irrelevant population on the political/economic stage.
Meanwhile domestically the entire economic model is failing and the “union” is getting shakey as its 2 biggest members turn more nationalist.
Maybe this “oh you cant compete but here’s a trophy for sovereignty” attitude isnt helping. Less clapping along with the EU bureaucracy’s latest make work projects like transitioning to a new Word processor. If we want the EU to succeed, more tough love is needed imo.
it's industrial policy, which is of course the minimum for being able to even think about having independent thoughs when it comes to international relations.
hard to stand up for EU (or even national) values if some dude in Washington has the keys for most of your military
having at least some in-house expertise is the first step.
I agree, but my point is I don’t think we should be cheering on a group of 600 million highly educated people struggling to achieve the bare minimum.
We should be shaming them into contributing something useful to the rest of world in the most important development of our current time (AI) instead of clapping for domestic regulatory moats.
They aren't struggling. They have generally outperformed all US open-weights model releases by a decent margin. OpenAI and Anthropic have no similar product.
From where I'm standing, as an American, we should be shaming OpenAI and Anthropic for both pushing towards an IPO while neither of them have profitable revenue streams. In pure game-theory terms, you could successfully argue that the US is the biggest loser of the AI race: an economy propped up on artificial Nvidia margins and terminally reliant on neoliberal trade.
Even if global war doesn't tear down the TSMC house of cards, neither the government, the banks, nor the global economy could underwrite an OpenAI IPO. They make no money!
Anthropic is profitable: https://www.forbes.com/sites/jonmarkman/2026/08/17/anthropic...
FTA: The same disclosure showed positive adjusted operating income for the period. The figures are preliminary, unaudited, and could be revised before the prospectus lands.
Like, adjusted can cover a whole host of sins. Until we see the GAAP audited numbers, nobody but Anthropic employees and investors know.
And still, analysts are calling it an even bigger gamble than WeWork's IPO: https://www.cnbc.com/2026/10/07/anthropic-will-be-most-ridic...
An economy “reliant on Neoliberal trade” just describes the entire global economy. Your quality of life goes back centuries in any other scenario.
The fact you’ve used the trendy populist boogieman word “neoliberal” and complained about a generational startup not being profitable during the growth phase tells me you watch a lot of internet content from people who don’t understand the topics they cover (breaking points perhaps?)
I would recommend re-evaluating who you borrow your opinions from. Global trade is not a boogieman, and OpenAI will have zero trouble building a profitable consumer advertising business, regardless of whether their IPO flops in the short term.
Ah, the current global economy. What about Russia? The goal was to liberalize post-USSR Russia... until they lost access to certain markets due to sanctions. Same for China, Iran, North Korea, they're all a shadow of what's to come.
I didn't make a mistake, or reach into the buzzword hat for that; Nvidia's trillion-dollar GPGPU products are wholly dependent on TSMC. Every single Blackwell die ever made came out of a Taiwanese factory, and the US leadership is inviting rapprochement with the CPP on the eve of Taiwan's invasion. TSMC will destroy their machines before China gets them, and then there are no more Blackwell products. US fabs do not ship a comparable fabrication node and likely won't for another ~5 years, optimistically speaking. This is not a "war will make the stocks tumble" argument, we are describing a scenario where American economic growth could slow or stop entirely if AI dominance is jeopardized. The US economy can't have an Achilles heel like that.
It gets worse! There is currently a global DRAM, SRAM, HBM and flash memory shortage that has already encouraged Samsung, SK Hynix and Micron to start fucking over their most loyal customers like Apple and Google, who have absolutely no alternative vendors. If you're packaging SOCs, neoliberalism already fucked you over! South Korea is counting beaucoup bucks for every iPhone sold at eye-watering prices. Taiwan is gouging US OEMs, making Apple and Nvidia engage in humiliating bidding contests to secure access to a single factory line. If global conflict expands, the AI economy will be chronically dependent on government (read: taxpayer) money. US taxpayers don't have trillions of apology dollars for this type of mistake.
China's got the edge on open models. Free trade does their heavy lifting, US labs are scrambling to put up defenses and beg for pathetic regulatory efforts that pull up the ladder. Yes, neoliberalism keeps them up at night, and apparently it keeps you up too.
If you think about it EU is under pressure from US, China and Russia. At the same time the EU politicians are hellbent on ideological energy policies that made EU extremely vulnerable to price swings. With all that wealth and economic growth EU should have been energy independent ages ago.
At the same time France is ablaze now by younger generation asking for a better future while the public spending is 57% of the GDP in 2025[0] and stuck in a vicious cycle.
Only Germany is holding the fiscal stability by a thread and that's even questionable given the current German government is more like a zombie government that cannot agree on the important reforms and most probably would collapse before the end of this year.
It's completely valid to question how "Sovereignty" helps EU when the complete backend (chips, energy, technology, etc.) is imported from US or China, Middle East and just slapping a Sovereign sticker at the frontend on a mediocre product that even most of the europeans don't want to use given the alternatives.
[0] https://tradingeconomics.com/france/government-spending-to-g...
Just ran this through our data analytics benchmark (I work at Plotly).
It's 10x cheaper than Mistral Medium 3.5 from April and goes from 58% to 74% correct. Definitely a generational shift.
It's not on the Pareto curve yet, but it's good enough for data analytics, and at this rate I suspect it'll be excellent in another few months.
Full write up: https://plotly.com/blog/mistral-large-4-plotly-data-analytic...
If I'm reading that plot correctly, qwen3.8-27b beats in accuracy and price?
That’s correct. About the same in price, but Qwen does get more answers correct in this particular benchmark.
See, that's the sad part. You're over a hundred more parameters, and are less than a hundred times better. I'm not sure how costs scale per parameter count, especially for moe vs dense models, but I honestly don't know when the economics will work for trillion param models when smaller ones are almost as good.
I know quality isn't strictly additive but really can a trillion param model beat even 10 clustered billion parameter models working together at the same time? This is the same horizontal and vertical scaling thing we have with normal systems… there's a reason why servers don't run at 5ghz like gaming rigs…
Id guess it depends: The simpler the task, the more useful to split it among many smaller and cheaper AIs. While the harder the task (eg unsolved maths problems), the more important with a big and bright model.
That's a big jump in the right direction for 6 months. It's certainly more on the level of the light models in terms of cost and performance, but the trajectory is awesome!
A strong competitor in cybersecurity as an alternative to GLM-5.3 (Mistral reports 82% on CyberGym-E2E). Visual grounding is also impressive (42% on Dense 200 vs. 41% for GPT-6 Astra).
Otherwise, behind on the broader Pareto frontier, but not by much (Vals Index: 48.05% vs. GLM-5.3’s 53.51%; $13.78 vs. $7.25 per test). Many companies will prefer it over Chinease models.
That was about my conclusion as well: Slightly less than GLM 5.3 performance but made in Europe. So, maybe it answers Tiananmen Square questions correctly, and in French. All in all, a reasonable model, but not frontier.
> So, maybe it answers Tiananmen Square questions correctly
What would you consider a "correct" answer? I just asked deepseek-v4.1-flash asking it what happened (without mentioning the word "protest"; here are some excerpts of what it said:
> In April 1989, students in Beijing began demonstrations after the death of Hu Yaobang, a former Communist Party general secretary. The protests grew. [..] Estimates from other sources range from hundreds to several thousand deaths. [..] The Chinese government describes the events as a counter-revolutionary riot and says the military action was necessary to restore stability. It restricts public discussion of the events inside China. Many other governments, human rights organizations, and observers describe the events as a violent suppression of peaceful protests.
So, let's see... it calls it a "protest", mentions the number of deaths, and even mentions the censorship of the topic by the CCP.
Do you not see how twisted that reporting is? It is doing the bare minimum to affirm that something happened, while injecting enough uncertainty to make it sound like it’s an issue blown out of proportion by Western media.
I didn't paste the full reply. Nevertheless, I can confirm the reply was factually correct and in no way seemed like it was deflecting and/or making it seem like an issue blown out of proportion.
Another example, if I ask it about the CCP and their wrongdoings it tells me flat out (I copy-paste):
The only twisted thing here is people's bias; because it's a Chinese model it must be censored and biased in a way that's pro CCP, right?Your quoted text states the government position that the protests were a "counter-revolutionary riot." It does not state the protester's goals, the reforms they were asking for, or any context that might counter this narrative. Only that "governments, human rights organizations, and observers" object to that description. HRO's and observers are paid for by western governments, and widely depicted as foreign agents / CIA fronts. This is exactly what you would expect a carefully crafted political statement to be, such that it reads entirely differently to different audiences.
Regardless of what it answered you would always be able to find a way to critique it. It's just a model. You didn't ask it any of those questions so why would it answer them? At least they answer the question and not "this is unsafe" like claude.
Have you compared it with the same prompt on openai or anthropic, and more importantly against other Chinese models?
What did you expect? Every time people repeat this "muh Chinese censorship" line they're consistently rebuked with results from Chinese models. What specifically is the problem?
It feels like we're shifting goalposts constantly: "Chinese models won't discuss Tiananmen" -> "They discuss it but lie" -> "They're twisting the truth" -> "There is no counterfactual information in this but it hurts my feelings" is where we seem to be now. Is it jingoist butthurt that makes people backpedal like this? Or something different?
I'm bewildered, and still do not know why people need AI to score perfectly on their prejudicial Tiananmen Square purity test.
Exactly, just like my reply above, people are only critiquing, it's super easy to mince words and pick out bones, but has anyone compared with western models with the same prompt? As long as it isn't dodging the question, it's good enough -- it even lists the kill count. If you wanted more details, ask it. If I ask a model "what is Google" I don't expect it to talk about the number of cancelled projects, privacy violations, etc. despite most people here would certainly love to hear about those things.
And then someone will come and flag me for being a CCP agent. Yes I have collected my social credit point, whatever the hell that is.
What is the problem?
> china bad
> eu bad
> murica best
rethoric and moving the goalpost meme
Meanwhile, most consumer facing Ai in the US refuse to discuss about our current president. (last time I prodded them)
I just tried, and ChatGPT gave me a very balanced view. Though it neglected to mention any of Trump's anti-democratic tendencies. I'm pretty sure this line of answering has been shoved in hard in post-training; it's walking a very delicate line.
> ChatGPT gave me a very balanced view. Though it neglected to mention any of Trump's anti-democratic tendencies
Do you hear yourself? How was it "very balanced" then?
> it's walking a very delicate line.
Perhaps even more delicate than the line DeepSeek and friends need to walk.
a good follow up.question would be to ask it about January 6, 2021
see if that is "very balanced" too lol
---
Gemini does not want to answer "was the 2020 US election stolen?" and replies with "you can check official sources" and then gives a bunch of government links
it did use the word "attack" when it described Jan 6
Its quite interesting to see that at least the early days of AI so far have not been a winner-take-all runaway acceleration game where catchup is impossible.
I certainly wouldnt have predicted that 10 years ago.
Very glad to see Mistral still in the game even after some big stumbles with Large 3. I deeply hope that this model is 'good enough' that it becomes the European go-to, giving them the resources to keep the pace up.
I'm excited to try this out today.
I think a big part of that is the Chinese publishing the solution for everywhere hurdle in the road they've encountered in the form of a paper.
Deepseek essentially releases instruction manuals in paper form.
I think it might have accelerated things but on a much more basic level, there seems to be no real moat in synthesizing the world’s knowledge into LLMs.
There's no question that training leading LLMs requires some serious expertise and know-how, but surely already having advanced LLMs/agents must be helping tremendously not only for software engineers but also for those working on LLMs themselves.
I think one could describe LLM optimization as "hard but not a moat". Years ago, optimizing neural nets was described "graduate student descent" - it's tricky but throw enough conventionally smart people at it and it will happen. It's like tuning a hot rod and finding a reproducible bug in a large code base. It's hard and there are tricks but not absolute hurdles, no problems waiting for a conceptual breakthrough (and at today's scales, are there any problems waiting for an Einstein to solve? That's an open (AI) question).
I also think we're seeing the sigmoid approaching.
I want that to be true (assuming you mean specifically the second half of the sigmoid) just to give me room to adapt to the changes we've already seen; but I've seen comments saying things are slowing down since around when GPT-4 came out.
What is really going on: all the AI labs are doing panicked model releases (and panicked training of new ones) because Qwen4 is rumored to come out end of October and is rumored to be very nice. Question is: is it another "Deepseek-moment" nice? Or just nice?
Btw: with Qwen4 I mean the next large Qwen model that is based on the Qwen4 architecture (Qwen 3.8 flash next was "almost" based on the new arch but obviously was a small model)
It doesn’t matter.
What matters more is if firm’s start using a bundle of American and Chinese models and when they find their feet - how large is the market for frontier?
Frontier has to displace labour one for one at some point or it’s over.
I'm pretty sure the point of Qwen3.8-Flash-Next was to get the open source engines to integrate the qwen4 architecture.
The fact that it basically broke open the local model supremacy was just a nice side effect.
I'm running: https://github.com/peonist-ai/halogen-server with a quant4, PLE offloaded, and it's resident VRAM is 36GB at 265k context.
Shave 10 more GB off and the TAM openai and anthropic are targeting is a lost cause. Local models are what 90% of people will need.
If the world governments can get a handle on the memory cartel, then there's no more moat for most normal humans.
I wondered if there would be a Qwen 4 or we would go straight to 5, re: tetraphobia, but perhaps it's more like an uno reverse card in this case
https://en.wikipedia.org/wiki/Tetraphobia
DeepSeek being Chinese also has 4 so I don't think it's a big deal for model makers.
GLM moved through their 4-series without consequence
I wonder if the next DS models will also graduate to 5.x, I think I saw they are training up a 10T model, and just raised $12B too
> and just raised $12B too
It's important to realise that Deepseek are a very unconventional model lab, as they were originally funded by the founder's work in quantitative trading.
They mostly don't need that much money, the reason they took funding is to provide employees with equity, as lots of their good people were getting poached by other Chinese model startups.
(I have no inside info, just read the FT articles about them).
Awesome 3.8 next runs great on my Framework Desktop so I'm loving more local models.
I have similar hardware - what specific version of the 3.8 Next model are you running and how many tokens/s are you seeing? 3.6 35B A3B gets about 68t/s for me so I've been sticking with that model for the mean time.
The thing with 3.8 next is that it uses a variant of ngrams. Part of the network is replaced by a lookup table you can store on a fast ssd.
In practice, you will be able to run models a bit bigger than 35B.
https://unsloth.ai/docs/models/qwen3.8-next
https://github.com/peonist-ai/halogen-server is beating the pants off of anything I've tried.
38GB of vram resident. more tk/s, more prefill.
I think the moat is going to be compute. So far compute needed to push the frontier is still extremely cheap so the capital can afford to spread its bets. But when further improvement is going to cost in trillions, capital will have to pick a winner and bet only on him. It won't be a matter of finding the best bet, it will be a matter of survival.
This will cause the picked winner to get massively ahead with sheer compute alone used both for training and inference dedicated to recursive self improvement.
I guess all predictions age like milk, but here's one:
There's a law of diminishing returns at play here, and doubling the energy cost of training to wring 2% more performance out of the technology isn't going to be very useful, because most of the problems it is capable of solving will be solvable with the previous-gen 98%-as-good model.
("there's a law of diminishing returns at play here" is an article of faith. But then, so is the belief that these models will keep getting better).
As soon as you can demonstrate decent financial returns (ie. the AI can run a company better than humans can), suddenly it makes sense to put a lot more $$$ in even if returns are diminishing - since whoever runs companies the best gets control of a big chunk of the world economy.
“ ie. the AI can run a company better than humans can”
lol You can always tell who has never ran a business before with comments like this
I understood their point, if we truly get AGI then no reason to think AI would be worse than a human.
AGI with people skills?
Chatbots are more considerate than many people I know. They at least explain things in a calm way.
Do you doubt? Already plenty of people prefer people skills of AI chatbots to those of other people.
If that works... why not just jump straight to a planned economy run by LLM? Skip the whole messy "free market" thing altogether?
(I don't think it will work).
What do you have against multi-agent reinforcement learning systems and why do you think they are not AI?
londons_explore is arguing for a winner-takes-all scenario, with an early advantage locking everyone and everything else out.
Surely there is a point where algorithmic improvements will be more cost effective than buying more hardware.
If you're actually applying LLMs, all of the things around the LLM that adapt it to coding, for example, that enable it to use existing validation tools for code, and enable it to diagnose and fix tool chain issues that aren't directly coding problems, are what makes the difference between a model that that scores a little higher on a coding benchmark and a model that's useful in a particular code base on a particular platform.
Are there any use cases that have enabled one customer of a frontier LLM to outperform a competitor using a different frontier LLM? Or is this why we are seeing confected points of comparison like solving challenge problems in mathematics?
I'm afraid it might be the other way around. RSI might pick all of the low hanging fruit soon. There must be a physical limit of how much intelligence you can squeeze out of some amount of parameters and compute.
There are going to still be worthwhile improvements but they are going to be more like not how to make transformers 10x cheaper but how to make next training run cost 9 trillions instead of 10 with a very particular optimization designed at the cost of hundreds of millions for this one specific run.
I think we’re no where near a physical information theoretic limit.
The hardware also is, so there ought to be a whole lot more room for improvement.
It is just occurring to me that “RSI” expands to recursive self improvement. Thought people were talking about repetitive stress injuries; either in regards to programmers writing too much code/not having to write code anymore, or the frontier AI companies and their tendency to applaud themselves.
But the old models still exist at trivial marginal cost. The frontier models would need to dominate every price point to really take all and so far they haven't been.
Don't worry, AI boosters will be in here soon denigrating anyone that uses anything but the latest and greatest models as irrelevant.
Not just compute but energy. Most of Europe has no access to the cost effective power generation needed
Europe has lots of zero-cost windows for electricity, and areas with cheap prices. The real issue is access to oil and gas.
Maybe for training big models one can wait for times when the wind is blowing.
There are worse ideas.
I could imagine a belt of data centres around the equator, that hand off their computational loads as the sun sets. Good scifi-esque premise.
That does not really sound practical to let datacenters costing billions idle half the time.
France is actually pretty cheap in Europe. About 15% more than average USA electric prices (but I know that varies a lot across the states so still likely much more than the cheaper areas)
Build Nuclear, Build Thorium the Chinese are building whatever they can. They’re not locked in by special interest. Is that because they have lots of engineers on the job in government?
Training location is flexible. Iceland?
My spend on DeepSeek is not much and I regularly top up my balance every month as my support for all the good work DeepSeek is doing for the open science.
DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government. The reason they release the AI models is economic warfare against US, not because of charity or kindness. It's great for us consumers, but the goal is not to help humanity or open-source.
Framing it that way hides all the levers used to tilt the playing field for US companies, no?
The government that removed restrictions on how private companies can access capital after a certain scale (the JOBS Act), that removed the need for private companies to report as if they were a public company after a shareholder threshold was crossed, superpowering the access of wealthy private investors to get in earlier in a growing company while at the same blocking the public from participating in funding growing enterprises at an earlier stage (since it required companies to IPO much earlier to access capital) which allowed retail investors to also reap the rewards on funding them early when they grew to become behemoths (like Amazon, Meta/Facebook, Google, etc.).
It's not fair in either place, the USA has its own model of unfairness, China has a completely different one. The difference is that in the USA the government allows private investors to become more powerful than the State (outside of the monopoly of violence) while plunging the rest of society into increasingly more precarious lives while in China the State is the power and its legitimacy only exists while the population feel they have a better life.
IMO the selective enforcement of regulatory requirements should be added to the list. Observing from the outside, I have a hard time believing that e.g. musks gas powered data centers really follow all the environmental laws, for example, or that the authorities really see no grounds for indictment if Altmans company hacks hundreds of third parties, or that there's really no one at the SEC having a problem with Anthropics fear mongering prior to the IPO.
In a list describing how the USA system works, sure: selective enforcement is there.
As a differentiator from the Chinese system, not so much. For two otherwise equal companies, the one that says things against the party line will experience selective enforcement too.
Except that in China, the people who do that will disappear.
The propaganda of trying to make US and China government appear the same is making people dumber. and is one of the most heavily used tools in China’s online propaganda arsenal.
In the USA, they don’t disappear, they commit suicide in their car typically. In Russia, they fall out of windows and off of balconies. I don’t think it’s accurate to call the circumstances equivalent, or the regimes equivalent. But there are some remarkable similarities, and their goals are similarly structured - to keep those in power powerful and to enrich the wealthy.
>The propaganda of trying to make US and China government appear the same [...] is one of the most heavily used tools in China’s online propaganda arsenal.
The government could do some very easy things to remove this tool from China's toolbelt. Like you know, prosecuting rich people when they've earned it.
No. They would need to kidnap rich and poor people and their families regardless of "earning" it. Then they. will start to get closer.
Who was trying to make the US and China appear the same? I think you have reading comprehension problems, buddy.
If you look at the comment I responded to, they were remarking on selective enforcement in the USA. I merely observed that this isn’t a unique feature of the US system, it happens in China too. I said nothing about the severity of the consequences.
he's not really "framing" it that way. that's just simply exactly what it is.
There's a framing though, by mentioning only Chinese tactics as nefarious/malicious it leaves the USA as the morally superior model (and for many it will be assumed as the "fair" option), which it definitely isn't.
If you think about it, the US labs are heavily subsidised too. Not only they receive billions in state funding, the administration is also prepared to engage in trade wars to help them.
I think the Chinese government is backing their labs by less direct means. For example cheap electricity and investing in chip manufacturers such as Huawei and cxmt.
Deepseek specifically, is known to operate with minimal resources. The entire company has around 160 employees and every model they release must break even within ten months.
SpaceX has been awarded roughly $22 billion to nearly $30 billion in cumulative public federal contracts, the $280 billion CHIP act, and who knows how much the CIA + NSA are spending.
Awarded that for doing work, though, not to clone someone else's work and release it for less money.
I think this is different for different labs. Some innovate while others copy.
Same goes for US labs, not everyone are as innovative as google deepmind:
https://www.forbes.com/sites/antoniopequenoiv/2026/04/30/elo...
> not to clone someone else's work
Ironically your comment will be cloned and used for training.
Whose work on KV-cache reduction did Deepskeep clone? Please be specific.
It seems obviously dustortionary to highlight the niche, optional part of the process they did spend time on, that would be utterly worthless without the difficult bit being done elsewhere. Please be less specific.
> not to clone someone else's work and release it for less money.
... You're saying this about _deepseek_?
> Awarded that for doing work, though
Yes, just that. Being a huge supporter of the regime - at a time running a department - surely has nothing to do with it.
Yes. If you think SpaceX is some sort of terrible also-ran that can only get contracts through political links, then honestly I don't think I can help you there. I don't think anyone can.
You see why there are so many conspiracy theories about US landing on the moon, one is left wondering how is it possible that a countries of such retards did something like that
I'm pretty sure, the training/fine tune through Claude/OpenAi would not be possible without the army/China's hacking teams (and legal protection). So it's more than just a cheap electricity.
Supporting deepseek is just like supporting the Chinese army, no need for that. Though it goes both ways, OpenAi subscription just lowers the cost of the US army as well.
Sorry, are you saying that it's a subsidy that it's legal for them to distill other models, or what? It's not even clear that there's any kind of protection for models in Western countries. Why would they be worried about the legality of distillation? And what does the army have to do with distilling models? Like you're saying it's a subsidy that China protects their borders from invasion by the US?
It's not a subsidy for the Chinese to say they're going to ignore our IP laws. It's a subsidy for us to say we're going to make them and try to push them on the world. It's literally granting a monopoly by legal force. It's very obviously not aligned with the interests of the American people, while China releasing things in the open is.
Same with the US legal system protecting Anthropic from copyright laws from all the books and other stuff used for pretraining...
>Supporting deepseek is just like supporting the Chinese army
Good, where do I sign up? At least they aren't exploding little children and generating chaos in the oil market.
https://www.business-humanrights.org/en/latest-news/anthropi...
I want you to come here and apologize in public when Xi starts bombing Taiwan like he's promising.
I want the current president who has lost his mind to apologize to two California cities.
You and me both. But he'll be irrelevant in two years, maybe less. Xi is in for life.
As bad as casualties get in a future war like that, it's still very likely to be a symmetric war where the main targets are military infrastructure. No systematic annihilation of universities, hospitals, religious sites, journalists and health workers like certain genocidal regimes backed by the tech oligarchies have been doing.
https://www.ohchr.org/en/press-releases/2025/06/israeli-atta...
https://www.unesco.org/en/articles/unesco-presents-assessmen...
I want you to publicly say that this turtle biologist and this journalist are Hamas: https://www.theguardian.com/world/2026/jun/20/mona-khalil-tu... https://www.theguardian.com/world/2026/apr/23/lebanon-journa...
When one player tries to monopolize AI tech and market by all means, others are those who don't want to be hooked to a foreign will in the future. I think this is the reason in doing open research, and it's likely positive for all of us.
> I think the Chinese government is backing their labs by less direct means.
It may be that the effect of the backing in both places is essentially equal but this statement is strange. The Chinese government invests directly in Deepseek[1]. Notice the article, in addition to saying the CCP is investing, says Tencent is also a major backer. CCP owns a golden share of Tencent.
[1]https://www.cnbc.com/2026/10/06/deepseek-funding-round.html
Ok, but Trump stated that the US is in active talks to take over parts of both OpenAI and Anthropic, too: https://www.cnbc.com/2026/06/05/trump-open-ai-altman-stake.h...
I'm not arguing about a comparison here. I'm trying to correct a misconception about China that I see a lot from Western perspectives.
Everything a Chinese company does has the explicit backing of the CCP, at least ideologically and usually financially in some fashion. While DeepSeek may not be the CCP, it couldn't exist if it was expressing any kind of ideology that wasn't inline with them and, in this case, is explicitly funded by them.
A Chinese company has no freedom to say Taiwan is a country the way someone in the US could suggest California succeed from the nation.
Any public message you hear coming out of China has the implicit approval of the Chinese government.
I would like to point out that a lawsuit is currently live about the us govt forcing US tech companies to retaliate against us citizen criticizing the govt’s policy.
Well, there was one happy audience in Nebraska with the blessing of El Presidente that was happy if two California cities get the shaft.
> Everything a Chinese company does has the explicit backing of the CCP, at least ideologically and usually financially in some fashion.
If you're not arguing about comparison then don't state this as if it's any different from the US. Because you make it sound that way. Or do you think that if tomorrow OpenAI came out vocally supporting the DSA - that suddenly they wouldn't see lots of barriers rise up out of nowhere?
Come on now. That's fairy tale land.
Retaliation is different than what happens in China. In China, the DeepSeek CEO drafts a press release about Taiwan's independence and he's promptly fired and sent to prison.
If OpenAI comes out in support of _any_ political party, they may lose business from the government in the future.
> If OpenAI comes out in support of _any_ political party, they may lose business from the government in the future.
Fairy tale land it is. Consequences for what I said would be far bigger than "lose business from government".
Well, they are helping 17% of the world's population, and the US is currently actively engaged in trade wars and economic warfare or explicitly attempting to leverage it's hegemony against its long term alliea for short term gain.
There is as much to criticize about American hyper scalers and AI labs and the lack of interest in helping humanity or contributing to open source, but that might not be as popular an opinions on this site.
Only 17%? I'd say driving down the price of AI helps at least 95% of the world's population. It certainly helps me.
I assume 17% is roughly the percentage of people that use LLMs globally (the numbers vary between 15 and 20 percent).
Amazingly low, given potential impact. Where is this use concentrated? Who has the best data on use right now.
17% is the approximate population of Earth in China. I was using a simple example as a counter to the anglocentric view the parent poster was expressing by trying to paint China as a villain.
> Well, they are helping 17% of the world's population,
Any open weights model that Chinese companies are releasing for free are allowing everyone in the world to have access to high quality realizations of these tools without the risk of the US regime arbitrarily cutting you off.
Also, it's funny how all the neoliberal mantras of how free market drives progress through competition stops when it's a US company that's being challenged.
I suppose in some sense preventing nuclear holocaust is a short term gain...
What nuclear holocaust has the only country to use a nuclear weapon against a target prevented?
Do you think the world is closer to or more distant from large scale global conflict today than it was 10 years ago, or even 3 years ago?
Iran has expressed a willingness/eagerness to nuke Israel and the US. That outcome would be less than ideal
> Iran has expressed a willingness/eagerness to nuke Israel and the US.
Source? The last Supreme Leader of Iran issued a fatwa against nukes, way before he and his extended family where bombed in their home in a failed attempt at a decapitation strike.
> The reason they release the AI models is economic warfare against US
The story is so much more complicated than that, to the point that this economic warfare theory is basically a meme.
Chinese models are open because they don’t have a choice. “When you trail the frontier, openness maximizes reputation per unit of capability. The moment you lead, you close.” [0]
[0] https://earnedintuition.substack.com/p/involution-without-ex...
They do have a choice, as evidenced by all the times when the other option gets picked. The article you link tries to acknowledge ByteDance's Doubao and Alibaba's Qwen Max as exceptions, but forgets about Baidu's Ernie and iFlytek's Spark, which are also closed-weight. There's simply no consensus yet on which strategy is better, so different companies end up making different bets.
That’s because Baidu sells model access and iFlytek sells integrated products. Also both have released open models, just older ones. They just close their latest ones because they got the reputation they needed out of them and have transitioned to a product. Notice that both of these explain the motivation without resorting to unfalsifiable claims about "economic warfare".
Huh, when XHToken released a model named Spark, I assumed it was just a name collision, since iFlytek appears nowhere in the documentation. But on iFlytek's model marketplace, it's listed as their self-developed model: https://maas.xfyun.cn/modelSquare So they actually released an open model quite recently! Thanks for letting me know.
So is American models. They are subsidised heavily but still can't provide cheaper access. Their fault is to assume all countries can afford them.
By whom?
If you want a direct example of the US subsidising a major AI lab, we can use Microsoft's investment in OpenAI. Microsoft has the largest ownership stake - 27%.
Microsoft did not pay with money - it paid (mostly) with Azure cloud computing credits. MSOFT is then able to write this off as a loss against tax.
It is generally far more tax-efficient in the US to write a loss in this way than it is to write a loss for a cash investment.
In this case, I believe the difference was ultimately highly significant. When including MSOFT eventually writing-off the deprecating Azure hardware it had used to buy the OpenAI equity, the result was MSOFT's tax reduction being either close-to or exceeding the actual cash value of MSOFT's investment in OpenAI - IIRC.
These examples represent taxes that the US chooses not to collect - the US could choose to make investments like these less tax-efficient. Instead, by making them extremely tax-efficient, the US subsidises the transaction hugely.
This is a joke right?
Even ignoring monetary subsidies, there are the non-monetary ones: not being sued into oblivious by the government for their countless hacks of other companies and countries, the slaps on the wrist for massive piracy, the waving of environmental (and other) regulations in order to allow their data centres to be built an operated.
> Even ignoring monetary subsidies
That's what I'm asking - which monetary subsidies?
> not being sued into oblivious by the government for their countless hacks of other companies and countries
That isn't normally how enforcement works, and it hasn't been very long since they disclosed those breaches. If the victims want to pursue legal action, they can, and they still may!
> the slaps on the wrist for massive piracy
So judges and juries are involved in the subsidization conspiracy, too?
> the waving of environmental (and other) regulations in order to allow their data centres to be built an operated
Sure, though if you think this isn't happening in China too, I have a bridge to sell you.
CHIPS act, OBBA, defense, AI upskill grants.
That's just the ones i know off the top of my head in the US. Those programs account for more than 60 billion in committed spend.
The original comments is "DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government"
> That's what I'm asking - which monetary subsidies?
I am listing for you the non-monetary subsidies, which are just as real and equally important.
> not being sued into oblivious by the government for their countless hacks of other companies and countries
If it was one of the Chinese labs doing this hacking, the government would be stepping in. If it was European labs they'd be stepping in. If it was you or I the government would be stepping in. That is a massive subsidy (they don't have to worry about the same legal fees and exposure) and being allowed to continue doing business is in fact priceless.
> So judges and juries are involved in the subsidization conspiracy, too?
There is no conspiracy. They are objectively operating by a different set of rules than you or I could operate in this market.
> Sure, though if you think this isn't happening in China too, I have a bridge to sell you.
I never said it wasn't, I'm saying that it's happening here and it's a very real subsidy.
> equally important
They really aren't. Nothing in the development of these systems, absolutely nothing, is as important as money. Stop these silly false equivalences.
These are not silly equivalences.
All the major labs would be dead in the water if the government acted on the things I mentioned. If they treated the labs the way they would have treated us for all those hacks. Or if they enforced pollution measures (or god forbid ban on-prem turbines because of the climate damage). Or rule that training on data is not fair use, or that ingesting GPL3 code and then turning that into weights counts as a derivative work. Or. Or. Or.
Those are all just as important as cash.
Nah, nothing here is a special advantage of US firms.
The hacks are pretty bland as long as there is no mens rea and the hacked victims do not want to press charges. This isn't a "subsidy", it's just our normal legal process at work.
The pollution question is winding its way through court systems and getting a lot of pushback. I wouldn't call this a subsidy, just a slow bureaucracy. And at any rate this isn't special to the US; certainly China's environmental policy is far more lenient.
With the training data, again there's nothing specific to US corporations here. Anyone can use that dataset. It's just how our legal system works. You might not like it, but it's the accepted law right now. This doesn't meet any reasonable definition of "subsidy".
On the other hand, shoveling billions of dollars of taxpayer funds into corporations is by definition a subsidy. Full stop.
Investors, but also the government in allowing these companies to siphon electricity away passing the increased cost to the consumers
>Investors, but also the government in allowing these companies to siphon electricity away passing the increased cost to the consumers
Not arbitrarily banning companies from buying a product from a supplier doesn't meet my definition of the term "subsidy".
Fair enough, but it sure meets the definition of 'helpful intervention'.
>Fair enough, but it sure meets the definition of 'helpful intervention'.
I don't see how. Anyone can buy electricity without any intervention from the government. An intervention is an action that changes what would happen by default.
By investors. OpenAI and Anthropic are not profitable (Anthropic is profitable if you allow them to invent what profitability means).
I heard someone calling the key metric in Anthropic financial reports EBBT: Earnings Before Bad Things[1] :-)
[1] Where "Bad Things" would be the typical interest, taxes, depreciation, amortisation plus the Anthropic specific employee compensation, LLM training (you know, for the LLM lab), revenue sharing agreements (which is a form of paying for infrastructure), etc.
A recent YouTube video by Patrick Boyle said the same thing. They are only profitable if you ignore all the costs that make them unprofitable such as paying employees and developing A models.
That's not how gross margin works.
For LLMs, marginal cost is just electricity
I made a statement about whether Anthropic was profitable I don't understand your reply.
Assuming you meant to reply to me and not someone else are you saying under GAAP accounting standards Anthropic is a profitable business because under GAAP accounting their only expense is electricity?
(I see what happened you skimmed the conversation and didn't follow what was being discussed.)
There's also things like training costs, spend on AI data centres and GPUs, R&D, sales and marketing, etc. See e.g. https://www.reuters.com/business/finance/anthropics-ipo-pros....
Those are not marginal cost or unit economics
If the "fixed costs" kill you unit economics won't save you. Scare quotes because true fixed costs generally have a long shelf life. Right now AI models die every 6-9 months.
1. And that marginal cost is large, much larger than the marginal cost of serving a Google Search query
2. The marginal cost is electricity until the hardware overflows. So it's continuous for electricity and a step function for hardware capacity.
Plus you don't address the core point. Frontier AI is capital intensive. A lot more capital intensive than regular software. The existing clouds supporting the entire internet (!!!!) were built for a fraction of the cost of AI infrastructure, and there isn't even an end in sight to this continuous hardware investment.
Let alone hardware refresh cycles.
AI is basically investment into roads.
Software used to be a magical place, closer to selling music albums but even better.
AI is a much worse business margin wise than regular software.
Investment is not a subsidy. Words still have meanings. Nobody in this thread has successfully answered the question: subsidized by whom?
Non-punishment is also not a subsidy; again, words have meanings. Let's use the correct words.
Indirect subsidies are an accepted economic concept and well studied. The posts you replied to are obviously talking about indirect subsidies and dismissing them because they didn't say "indirect" everywhere is just sophism.
But some investments are subsidized (tax liabilities) among all the other things.
The government, specifically Trump's government and the current money circle in AI inflating American company stocks. In addition to all that American models do not share their papers like Deepseek and Qwen do. So you can literally say Chinese models are doing it for charity at this point.
Of course you can say that, but you literally cannot be serious if you do!
Can you say a US company donating to a non-profit and writing off taxes on the basis of it is donating to charity?
Would you say you could be serious in stating this?
Sure, but I couldn't say that a US company writing about some work they did in paper form is charity. At least not seriously! If it were charity, Google might be the most charitable company of all time, since their researchers published "Attention Is All You Need." But it is not.
Google has published a lot of open source software and a lot of research papers: I do believe they used to be very charitable even if they are less so today.
They have other issues as well, but among other things, I'd classify them as somewhat charitable too.
Oh I am serious!
The US regime, which includes the ruling class: the US VCs and megacorps are just as much of an extension of the US regime. They're incredibly intertwined.
Capitalistic exploitation against the working classes.
> DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
DeepSeek is basically a research lab founded by a hedge fund guy, more than anything else.
> The reason they release the AI models is economic warfare against US, not because of charity or kindness
There are many other reasons Chinese companies releasing models open-source or open-weight makes strategic sense.
A really easy-to-understand example is a company who has a near-monopoly on "serving video content" releasing a video model openly.
If you can be relatively certain that video content created by a model (which you have trained, using data from your own platform) will be ultimately served on your own platform, thus generating revenue from watch-hours, it makes sense to make those models as widely-available as possible.
It's also a net-positive if people use your public research to build better video models, because - again - you are reasonably certain that the even-better content those new models produce will be watched on your platform.
The alternative would making models harder to access and learn from (broadly, the current western model). Many would argue that Google, in choosing to not optimise its video generation models for "availability", is directly causing less content to be uploaded to YouTube. This is the trade-off.
I don't know much about DeepSeek's financing specifically, which obviously doesn't release video models - so I don't know how directly this analogy runs, or who directly benefits from the extremely evident rising tide that the public release of DeepSeek's research creates. However, this does not negate the broader rising-tide effect of the scientific method.
It's certainly also true that it's geopolitically beneficial to be able to undercut American labs' models. If I ran a global superpower, I would probably want my country to be technologically competitive too.
But Chinese companies are already serving a huge volume of customers in a complex, existing marketplace, before even thinking about the US market, and it's overly simplistic to assume that their entire strategy revolves around economic warfare directed specifically at the US. It's more nuanced than that.
This is, of course, without even getting into opening the can-of-worms around whether US economic policy also results in the US state functionally subsidising technological innovation, how comparable that is to China's model, etc.
Interestingly if Google did release readily available video models then they'd likely be dealing with an order of magnitude of scale in the same way GitHub has had to.
Causing less content to be uploaded, when you are clearly the gorilla in the room, looks to be a wise strategy not a trade off.
I'm not sure I follow - do you mean an order of magnitude of compute demand?
Releasing models openly doesn't mean providing the compute. It's beneficial for relatively GPU-poor Chinese orgs to let users run models themselves - by releasing the models openly - as it means the orgs don't have to provide the compute to run their own models.
I'm not sure "subsidised" is the right word. If a government funds research and the results are released openly, that's just publicly funded research. It's how a lot of science works in the US and Europe too.
The open-weight models are a great boon to all American companies other than a handfull of Mega Corps in the A.I business. Care to explain how this is "economic warfare against the U.S"
Do you not see the risk inherit in sourcing everything from a antagonist?
Yes, hence why I'm happy there's competition so we're not stuck with US tech. After all, you've proven to be very unreliable allies to us.
and you've proven to be parasitic and free riding allies in return
Yes we see that in all other countries trying to avoid dependence on the US for everything. What we don’t see is competition with other countries having anything to do with warfare.
Economic warfare against the US? I mean maybe against specific US companies and stakeholders but on the whole it seems like it's good for the US economy as well as the rest of the world, kind of like supplying free electricity would be
Same goes for the US companies as well. Current US administration wants US to win the AI race at any cost and China is the only competitor left in the race. Winners always write the history or in this case the future of humanity
And OpenAI and Anthropic are just in for the love of the game? Our of sheer desire to help humanity?
Obviously not but they employ American researchers and aren't beholden to the CCP
It's one of these rare cases where intention is not what is the most important - the net benefit for consumers and companies outside of the USA is indisputable.
Not sure why so many people will vehemently refuse this idea. I won’t say it’s 100% true but it would be foolish to dismiss it. China is very much an adversary to America and has made it pretty clear they want to be a dominant leader of not the new world leader. Not here to evaluate what is good or bad. Keep in mind historically China has aggressively fostered industry (not unlike the west) but sometimes even more aggressively.
Americans need to travel to China. The Chinese have zero issues with us. They quite like Americans. This is such a weird propagandist take. Idk if you remember but both country's leaders just had a slumber party for 3 days in DC. This is not what enemies do.
Not even our leaders say China is an enemy, ita mostly businessmen who are scared of competition and trying to regulate chinese out of their markets so they can make more money milking us.
I have been to China many times. The citizens and visiting the country has nothing to do with global politics. Your take is just as propagandist as any other. China is not unlike the US and China has also made it subtly clear their desire to be a dominant global force. I am not marking judgement on it but don’t be a fool thinking that China is not some level of threat. This is how these discussions become weird. I don’t think China is going to blow up the US and I think it’s often a weird talking point by some US politicians but I also don’t think they are a peaceful actor but folks like you will suggest I am repeating propaganda. No, just pointing out that if you look at their language and actions they are an actor to pay attention to. Just because leaders meet means nothing.
The same is true of the Iranian people. There's sometimes a difference between the government and their own citizens.
Your just saying it’s beating america at its own game and crying fowl.
When did I cry fowl? I am simply saying that the truth is probably somewhere in the middle.
Which is the best kind of reason.
It is great for everyone except for a few people who want power over everyone else, and the fact it is not charity of kindness makes it more sustainable, because charity and kindness is quick to go when big money and politics is involved.
I want more warfare like this. Building stuff instead of destroying stuff.
They are also releasing the weights so I am more inclined to think they are better than most.
That would have been a scathing criticism if anyone believed any of the American companies have the goal of helping humanity or open-source.
Notwithstanding that Chinese publishing methods actually does help both humanity and open-source.
What about cheap solar panels and electric cars, is that economic warfare too?
To say that US model providers have a close relationship with their government would be putting it mildly.
At least DeepSeek didn't build its models on government subsidies. It came out of a quantitative hedge fund, so funding was never really an issue for them.
It just so happens that helping open source is the result. Image how far behind we’d (the hackers, not the moneymen) be as a sharing community be without them.
Whatever their reason, I'm glad they do it. This tech shouldn't be monopolised by a handful of closed, profit-seeking corporations.
And likewise, it shouldn't be monopolized by the CCP
What do you think your comment is refuting!!? Chinese are publishing their research publicly and opening their models! How on Earth would they monopolize anything?? Such nonsense. If and when they start copying the Americans and closing everything down, you can say stuff like that. Until then, ffs stop this nonsense.
Please look up predatory pricing. The only nonsense is assuming that publishing their research publicly and opening their models is altruistic
It doesn't matter if they are doing it out of altruism, the result benefits everyone anyway. It feels silly to have to explain this kind of thing. Are you trolling?
Well, US models are economic warfare too, of course.
> The reason they release the AI models is economic warfare against US, not because of charity or kindness. It's great for us consumers, but the goal is not to help humanity or open-source.
You put it like US has a goal of help humanity or open-source
If your belief is accurate, we should expect China to short the IPOs of Anthropic and OpenAI and release better frontier models immediately after their IPOs.
Does anyone think that likely? I have no clue or bias.
I don't think that's likely because my understanding is that Chinese people in mainland China have a tricky time shorting American stocks due to Chinese capital controls.
Economic warfare against the US is charity and kindness to a sizable portion of the world's population, especially when the US uses it's global hegemony as warfare against them. Neither system is perfect, but lets not be disingenuous.
So you’re telling me warfare doesn’t having to be blowing up schools and children, skyrocketing prices, crippling sanctions, and all that shit? Can I sign up for more of this warfare.
I’m neither in the camp of Chinese or the Americans(collective West) in general..
As a neutral party, this characterization is crazy.
As if the AI companies - Claude and OpenAI are guardians of freedom and humanity and very charitable to the global society without any self interests… “Chinese models are subsidized by the Chinese Government, therefore they’re inherently bad for humanity” is a highly propagandist argument. The politics of US vs China may be whatever it is in reality.. You have one company releasing their models for cheap, actually open sourcing their trained weights, and publishing details of their optimizations and learnings for others to use. The other camp actively “aligning” their models, nerfing their capabilities, hyping their swarm activities from poor sandboxes, and trying their best to lock users into their harnesses and walled platforms. They are subsidized by the capitalist VCs who are essentially waiting for their payouts..
At some point, one has to see things for what they are and evaluate their own reasoning..
I’m happy to stay provider agnostic, try all models and cheer any useful progress as open as possible.
Claude, and other American models, are heavily subsidized by investors (and the United States government).
If the means of achieving their goal are sometimes mistaken for charity and kindness... are they the good guys?
To be completely fair, write this kind of paragraph for every AI category. Would love to see your take.
I like when bad intentions produce good results, I'm tired of seeing the opposite in practice.
At some point though _the purpose of a system is what it does_, regardless of their intent.
How do you know? You’re not speaking for the Chinese government, are you?
Precisely. It's similar to their practice of aluminum dumping to depress US aluminum prices, which causes our aluminum mines and mills to close.
> DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
Yes, because everything China does is against the US. That's all they think about day and night. God forbid they want to corner the global market or have a genuine business case. How dare they provide options for those who can't afford a measly $200 a month? How can we let Chinese labs publish research for free for the whole world so that they can benefit? The nerve! To think they can use soft power instead of military might! I mean, Anthropic and OpenAI are the last bastions of human kindness and charity. Right?
Right?
Poe's Law applies here.
I wish US was rich enough to subsidise development of open source science and useful open AI models. I wouldn't mind US waging this kind of wconomic warefare against China or everyone else on the world.
Charity and kindness is not a motivation, it's an outcome of what you do.
> are heavily subsidised by the Chinese government
We hear this about literally every industry the Chinese excel in - that it's only because the government subsidizes them that they succeed. For chip manufacturing, for batteries, for EVs, for solar, for AI. I don't see how the chinese government can afford to subsidize all of these industries and still have them contribute to the GDP.
> DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
So were Amazon and Uber by the US, which have now established monopolies across the globe. To the countries suffering from those, there's zero difference with China doing it to solar. Actually there is, at least solar got them cheap renewable energy in return. This would never have happened in the US because big oil interests would make it take decades. That's the reality.
You need to spend 10 years outside the US, deprogram, and then go back.
> DeepSeek, and other Chinese models, are heavily subsidised by the Chinese government.
US companies are burning colossal piles of cash in ways that makes it unclear if it qualifies as dumping, not to mention their deep ties with the country's regime.
Claiming that companies from a country have ties to the regime and burn through cash is a very miopic accusation.
Meanwhile, American AI models are heavily subsidized by stock market speculation. Ultimately, the subsidies from both countries are flowing out of the pockets of individuals.
That's an interesting definition of subsidized. I don't thing I've ever heard it used like that before.
Ha! 10-15 years ago, I wouldn't have used it that way either. But the stock markets these days just feel entirely speculative. It is not clear to me if individuals are buying these AI stocks based on their expected future revenue/profit or the whole space is just momentum plays and the funds are directly being transferred back to the companies (by way of incentive packages and ability to get loans based on market valuations) to keep the train moving, despite there being no clear profitable destination for any of the LLM sellers. The hardware folks are making out like bandits, though.
US is basically doing economic warfare and bullying against everyone else ATM :)
What if economic warfare against the US does help humanity?
Why are we assuming a strong US is necessarily good? As a European, I have seen plenty of evidence against that stance lately.
I understand that Americans might prefer a strong US. But conflating them with humanity is a leap that I don't think one can make without any backing.
What evidence is that? Please share.
A conspiracy to make the US look bad by being better at producing all the goods and services the world needs at a reasonable price. Have they no shame?
Not really this is an X algo conspiracy. Up until recently the Chinese government wasn't even that invested in these companies. We're talking very very small grants compared to training costs.
Its very xenophobic of you to say China has zero intention of helping humanity, and just wants to "wage economic warfare".
Last time I checked, it was ourselves (USA) waging economic warfare on 2/3rds of the world.
I dont get this cope people have where people have this idea that its impossible for a Chinese company (that make billions of dollars) to have done something by their own merit, but instead its always some Chinese Communist Party conspiracy where the main goal is to destroy America.
Lay off twitter for a bit.
>Up until recently the Chinese government wasn't even that invested in these companies
>Last time I checked, it was ourselves (USA) waging economic warefare on 2/3rds of the world.
Up until recently the USA Wasn't waging economic "warefare" on 2/3rds of the world
The USA has been using financial sanctions, aka, financial warfare on anyone it's deemed an enemy for going on 4 decades.
China has been owning and controlling key companies in its industry for going on 6 decades.
Every country does this. Do you think the US government doesnt fund, regulate and control key companies?
China is not a threat to you, or anyone in the West.
What's wrong with governments subsidizing scientific work?
You think extremely US-centric. China has a different economic model than US and your rules for a specific kind of Capitalism may not apply to them. You assume a country of 1.4 Billion people is obsessed with a couple of foreign AI companies. What if they don't care.
> but the goal is not to help humanity or open-source
Ok, but so what? That's what happening so far. Even a repressive totalitarian government I wouldn't wish on my worst enemy does some good sometimes.
If the Chinese cheap/open models rise up and destroy us, thats on us for giving them access to the tools to do so.
Why the anti China propaganda from you? The United States is doing the same thing here and making American multi-billionaires even wealthier. The greed of American AI, GPU and memory and storage corporations is a black hole on availability to humans around the World. The result is a massive financial bubble promoted by the American Government to the detriment of our citizens. The USA has an AI ponzy scheme shuffling the same money between data center owners (Oracle & X), Nvidia and memory & storage vendors.
They're trying to pull digitally what they already pulled physically. The reason we can't manufacture a grill brush for a reasonable price is the result of years of Americans choosing the cheapest price. We gave up our manufacture base. They want us to give up our labs.
Deindustrializing a country is not something consumers can achieve. It starts at the top level, with politicians who construct a financial system where it's more profitable to speculate than to build or invest in real businesses.
Except they are open with the tech which is easy to replicate. All new models lowering cache prices is the result of DeepSeek's publications.
US economy is driven on quarter to quarter short-sightedness of stock gamblers and CEOs that have to secure their bonuses and perks.
Its them who chose to outsource heavily, not US consumers.
There are no victims here however, both benefited from the arrangement. The consumers and capitalists.
> The reason we can't manufacture a grill brush for a reasonable price is the result of years of Americans choosing the cheapest price.
No, it's the result of US leadership letting this happen. This is clear since China themselves would not let this happen. US leadership did nothing because they were best friends with the people who did it, and did not care one bit about their population.
Architectural/algorithmic tweaks do advance the efficiency frontier nicely. But raw intelligence mostly comes from data (not just its sheer quantity, but also how it's curated & cleansed) and the scaling law. The know-how about data curation doesn't seem to get published much, even among the open-weight labs, though.
This. Even in the efficiency frontier, it is a lot of data curation that actually makes many of those tweaks actually work at scale in practice.
So boring to see conversations moved over to Chinese models when that’s not even what we’re talking about here. This is about Mistral.
Europeans are irrelevant these days, surpassed by China, S Korea, Japan, Hong Kong, Singapore, etc. Europe is coasting on former glory and now has regulated itself to death and vacationed its advantages away.
Evidently not given that this model is on the open source frontier.
cost is higher than Chinese models that are better
I'm not sure I like this framing - so much of AI research has been academic, in the open, building on others people's work. Much less comp sci generally, math & philosophy, etc. The idea that rich companies can just build stuff in secret because they have resources is a fantasy.
Also, the field moves fast, but slower than people do. Researchers and engineers switch companies every year or two, and the know-how walks out the door with them.
There would be a lot of competition even without DeepSeek. Workers can freely exfiltrate trade secrets without noncompetes in California.
Proprietary competition, yes.
How it should be. Knowledge should not be copyrighted. The world will be a better place with such information democratized
>instruction manuals in paper form
So the most common way to publish manuals?
In research paper form.
I think the mean 'paper' in the scientific journal meaning; these are unfortunately often extremely bad 'instruction manuals'.
Maybe 30 years ago
> have not been a winner-take-all runaway acceleration game where catchup is impossible
From the Mistral site:
> ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe.
It is pretty capital intensive!
That cluster is literally orders of magnitude smaller than the compute pools used by Anthropic or OpenAI.
For training or for inference?
They don't publish numbers, but Anthropic has a single DC with 200k+ GPUs for inference, GPT-6 Astra is said to have trained on 100k+ GPUs.
Both, especially for training. Astra and Fable were presumably trained on cluster of 100,000k GPUs, or at least a couple of 10Ks.
3,800 GPUs is nothing in the frontier side.
I’m pretty impressed that they managed to get that close to the frontier with such a small cluster!
> I’m pretty impressed that they managed to get that close to the frontier with such a small cluster
Chinese companies also managed to put together their models with relatively small clusters.
Perhaps US companies are desperately trying to brute force their way into workable models?
According to Grok thats 7-10 MW. Tiny numbers.
To put that into context, the last wave of capacity SpaceXAI added 400-450 MW.
But how much of that are they using for training versus inference? They're serving quite a large user base.
These cards are like $3k each? That's, what, $12M and you keep the hardware? Honestly doesn't seem too bad.
More like $30k each.
Oh the server chip is 10x. That makes a lot more sense.
That’s kinda very small and light for modern trillion-param LLMs.
It's shaping up to be much more like a game of 'chicken' where each company tries to raise more cash without going bust... Ultimately the game of musical chairs is going to have to stop. In the US it looks like they are trying to get a government sanctioned truce in the form of regulation. That's what 'Pacing the frontier' means...
Am I reading this correctly?
This appears to be roughly as good as Sol 6.1 (which is quite good), considerably faster in terms of wall clock for complete tasks, and considerably cheaper (where Sol 6.1 is already good value - just really slow).
That seems too good to be true...
But I really hope it is true...
I can confirm that you're reading this incorrectly. There's a reason behind them only comparing it to open-source models released months ago. Here's a good aggregator: https://artificialanalysis.ai/#intelligence
Okay, so instead it's just another slower more expensive DeepSeek v4.1 Flash.
That seems much more realistic.
Even "runaway acceleration" isn't instantaneous. People imagine the singularity as something that happens almost instantaneously. But obviously it happens over time, and that time might be decades. It might still end up looking like a vertical line on a long-term graph.
If the singularity is defined as an AI sufficiently intelligent to improve itself independently, that AI is still limited by the resources required to do this improvement.
Hear! Hear! I really want European models / AI labs to succeed.
I trust them and their populations to provide a more societal-friendly version of AI, putting pressure on the US tech oligarchy, while also providing democracy-friendly open models that I don't trust to happen with the Chinese labs.
For sure people who don't grasp the difference between models, might be stuck in 'good enough' models.
But Opus 5.5/GPT is such a game changer in comparison to sooo many others, its still a moat for now.
You've got to think the comments about "good enough" are people who have not yet tried Opus 5.5. I haven't been this struck by a step change since 4.5/4.6. It's a bigger jump even than when Fable first arrived.
As for Mistral - I got really excited when they said Large 4 was focusing on being #1 in cybersecurity, because that's somewhere that they genuinely could edge out Anthropic & OpenAI. Have it actually solve problems, instead of Anthropic flagging "you tried to find a null pointer exception bug in your own code, we're now reporting you to the US government". But on the Mistral benchmarks I'm seeing, this looks very disappointing, but at least they haven't entirely given up. I genuinely thought Mistral had given up on new general models. They need to learn the bitter lesson all over again.
I feel like the "good enough" argument isn't about how big the gap between models is but about how good they are at solving the tasks at hand.
The capabilities of all models increasing so much all the time means there are simply less and less tasks you need a frontier model for.
Even if Opus 5.5 is 500x better than Deepseek, if deepseek can solve all my problems, why do I need to pay for more?
Many on HN still have the opinion that you must understand every line of code in the project, and that all is lost should you merge code that wasn't reviewed.
Obviously any model will do if you use it as a better autocomplete.
I believe that there is a large gap in expectations between different workflows.
Until the AI like reads my mind and produces perfectly production ready apps with minimal intervention from my side, there is still going to be room for improvement.
How I feel it Opus 5.5 and Astra 6 are both quite disappointing. They still write shit Rust. And they are very slow.
Now my provider serves DeepSeek around 300 tokens a second. The code is shit but I can have a few more rounds of corrections.
DeepSeek was maybe a dollar for the full PR. Opus/Astra over 50 dollars. And double that if you use their fast variants which matches the DeepSeek speeds on certain providers.
Yes, it is so good I rarely test new models anymore. Or think about cost.
The Rust I’m getting from Opus 5.5 works great. What’s the complaint?
Actually I felt since 4.5, it was “good enough” for basic coding (and competing models have reached that level as well). The big improvements since then have been longer-horizon tasks. Giving it a vague description of a problem, then come back hours later and find it did a better job than I would have. And since 5.5 I haven’t encountered any more load bearing seams.
Maybe it’s slow, but I’m not interacting with the models in a live chat anymore for most work. I’m giving them tasks to do async. The limiting factor is actually my bandwidth reviewing their output.
> You're absolutely right, I should not use a global mutex.
I am eyeing a huge codebase built with Opus. Did multiple passes of refactoring for things like using traits not just free functions. And using the Rust types like PathBuf from the damn standard library not just strings everywhere.
Oh and it is so hard to use a crate. Let's just vibe an inferior version and push another 5k line PR. Nobody reads them anyway right? Logos schlogos, a lexer is easy to write, right?
Maybe I am just getting old, but I remember a time when we were proud on the code we wrote. Now I read some of the old stuff and get a bit sad for what we lost.
Don't get me wrong DeepSeek also does this. But if the model is 100x more expensive I expect better. Of course if you never read the code it's ok. But it does the same mistakes and cannot design great architecture...
Hmm, I wonder if the difference is that I’m working on an existing large codebase that wasn’t originally written with AI. It might be picking up practices from the surrounding code. It’s happily using the same crates that are already there, and occasionally bringing in new ones. Some of the code is “weird” but I also learn from it.
> Many on HN still have the opinion that you must understand every line of code in the project, and that all is lost should you merge code that wasn't reviewed.
Uhm, yes?
Last week I elicited a parametric 3-D design model of a sand battery, a simplified thermal circuit model of it for rough exploration, a backward Euler multiphysics solver to evaluate its performance more accurately, and a volumetric visualization for the resulting temperature distributions, including some custom WebGL shaders: http://canonical.org/~kragen/sw/sandbattery/
It's about 12000 lines of code. I could probably understand every line of it if I spent a couple of months on it, although I'd need to learn some things about WebGL, numerical computation, and solid-state physics. But I elicited it in four days. Probably I'd be better off spending the next couple of months doing something else instead.
So, I was faced with the question of how I could keep the artifact thus created from being completely valueless. My solution was to export the heat-equation solution produced by the solver as (gzipped!) CSV, so that I can use my own code (that I do understand every line of!) to check that the solutions found by the solver are, in fact, solutions.
If that check checks out, the backward Euler solver may be of some value even if I don't understand every line of it.
If Opus 5.5 is 500x better than Deepseek, but Deepseek can solve all your problems, maybe you need to work on better problems. If you don't, and you're in business, your competitors will work on the better problems. If you're an employee, your employer might prefer to pay Anthropic instead of you. If you're doing projects you're interested in, you can tackle more ambitious projects with a more capable model.
This morning I elicited a microkernel operating system from Opus 5.5. Well, mostly. It doesn't implement task switching yet; we'll see if it runs into a wall at some point. But it boots in QEMU, and it's running a user process in ring 3 and serving web pages.
> maybe you need to work on better problems
I have enough real problems in life. I don't need to invent new ones just because a new technology is available.
Many of my problems in life are fully solved far past my satiation point by a 3b model that costs me nothing to run.
Many others are not.
But in either case, when I am acting and living wisely, almost all of my problems exist prior to the existence of technological solutions to those problems.
This is also true for the customers and employers that I care to work with. This has changed in me over time, but I now try my best to avoid inventing new problems. The world has enough big, important problems already.
> This morning I elicited a microkernel operating system from Opus 5.5.
This is cool but also a good example. I don't need a personalized microkernel just because it's possible to have one.
Maybe I need one and I don't know it, but the problem statement definitely isn't "I have inherent desire for a personalized microkernel".
You've misunderstood what I was talking about; that's a different meaning of the word "problem". Possibly you did not intend to start a merely semantic argument, but that's what you ended up doing, so I am unfortunately going to have to point at the dictionary.
You're talking about definition 1 in https://en.wiktionary.org/wiki/problem, "A difficulty that has to be resolved or dealt with," with the examples given being racism, addictions, and lack of access to health care. Those aren't the kind of problems AI can help with.
The kind of "problem" that AI can help with is definition 2, "A question to be answered, schoolwork exercise." That is the character of engineering "problems", although they are more open-ended than schoolwork exercises, because there are many defensible tradeoffs. "How can I build a bridge here?" or "How can I improve the fuel efficiency of this vehicle?" is a "problem" in the sense of a question to be answered with knowledge and effort, not in the sense of being similar to racism or addiction.
If the questions you're thinking of are so easy to answer that they can be easily answered by a hypothetical AI model 1/500th as good as Opus 5.5 — well, think up some harder questions.
Thanks for clarfying.
> well, think up some harder questions.
There's nothing wrong with this mindset, but your insistence that other people must have your values is off-putting.
To wit: at the end of my phd I was offered several tenure-track positions where "work on problems that are hard" would've been a big part of my job.
I turned those offers down due to my distaste with "because it's hard" as a justification for working on problems.
I'm much more interested in helping people and solving (def'n 1) problems. Those problems are almost always too large and messy to one-shot, but very often contain many components that can be solved with non-sota models, and make sense to solve with non-sota models for any number of economic/scaling reasons. E.g.: "can I serve this thing that costs $0.25 per run on even c.2024 sota models to 100,000 K12 students who will use it dozens of times a day on a non-profit's budget of ~$0.00."
P.S. The "problem" definition thing is very confusing to your readers because the person you responded to was clearly using the first definition! It would have been helpful for you to sign-post that you are using the same word as the person you are replying to but changing the meaning out from under your reader.
The person I responded to said:
I feel like the "good enough" argument isn't about how big the gap between models is but about how good they are at solving the tasks at hand.
The capabilities of all models increasing so much all the time means there are simply less and less tasks you need a frontier model for.
Even if Opus 5.5 is 500x better than Deepseek, if deepseek can solve all my problems, why do I need to pay for more?
They are clearly talking about "problems" that DeepSeek can solve, and they seem to be using the term "task" as a synonym for "problem", even using the problem verb "solve" with "tasks" as its object. To me, that makes it clear that they are not talking about problem₁s such as racism, addiction, and lack of access to health insurance; while you could imagine "solving" such problems, they are not tasks, and DeepSeek certainly cannot solve them. Rather, I think they are talking about "engineering problems", which are examples of problem₂, more analogous to math questions (to be diligently sought out) than to troubles (to be avoided).
It is certainly the case that the proposition that such problems should be diligently sought out is a value judgment, and therefore it derives from the value system employed by me, the writer. Would you prefer that I refrain from making normative statements in general, or just those with which you might disagree? In either case, is your preference specific to me personally, or do you consider it a duty incumbent upon everyone?
— ⁂ —
The 01913 edition of Webster's dictionary offers:
1. A question proposed for solution; a matter stated for examination or proof; hence, a matter difficult of solution or settlement; a doubtful case; a question involving doubt. --Bacon.
2. (Math.) Anything which is required to be done; as, in geometry, to bisect a line, to draw a perpendicular; or, in algebra, to find an unknown quantity.
Note: Problem differs from theorem in this, that a problem is something to be done, as to bisect a triangle, to describe a circle, etc.; a theorem is something to be proved, as that all the angles of a triangle are equal to two right angles.
The absence of the "trouble" definition suggests that perhaps it is derived from the "question" definition, and, surprisingly, very recently.
You gotta problem wit that?
The best problems to work on are not necessarily the hardest ones, nor the ones that need the most intelligence. They're the problems you, or other people, actually have. Are you going to give up on painting your deck because it's too easy and you don't need a 500x genius to do it?
Yeah no, if you elicit Opus 5.5 , anyone else can, and you have no moat either.
But if on the other hand, I mostly use my human intelligence and just need a dumb model to complement my human intelligence at low cost and high speed (say review every commit to catch obvious bugs), I have a much better chance of building an actual moat than you do.
But outside of coding, it’s even more clear that you don’t need frontier intelligence. My customer service agent is very happy with a 100B param Deepseek flash model, thank you!
> say review every commit to catch obvious bugs
I’m using subscription models for exactly that, better models catch more subtle bugs, and they catch them faster. It works out far better in terms of work-hours saved.
Also A/ then OAI slashed token pricing by 2x~5x on their latest models
Yes, obviously a microkernel operating system that you can vibecode in a morning is not a salable product. But it's still a perfectly good operating system, and sometimes people need those, and they're a huge pain in the ass to write, especially to debug. But sometimes off-the-shelf OSes aren't good enough. Something like 50% of the embedded operating system market is still "Other/Custom".
I use Opus 5.5 daily for my job. I am aware (and in awe of) it's capabilities.
Look at the context in which I used that term 'good enough'.
What i was saying is that there are tasks for which a dumber model can be good enough, and for organizations with sovereignty/ privacy concerns, those concerns can be strong enough to incentivize the use of a dumber model.
> I haven't been this struck by a step change since 4.5/4.6. It's a bigger jump even than when Fable first arrived.
I had the exact same experience. And unlike Fable, it doesn't gobble up your entire usage limit in a few hours.
I always wonder what the "good enough" people are actually using it for.
“Good enough” as in we don’t see any point in 1-shotting everything we want to build in lightning speed. If Opus 5.5 can 1 shot it then that product is essentially commodified, no point in any one building it except as an internal tool.
If Opus can’t 1-shot it, then it must rely on our human intelligence which can be complemented well enough with a dumb model as a frontier model.
I assume that you'd also argue in favour of working in a team with an average IQ of 80 as opposed to 120.
The thing is, for how long? Are they going to keep giving you "so much intelligence" for a "small" subscription dollar amount? When they really, for real, need to start making money to cover their costs, what do you think it's going to happen? Suddenly you will start having tasks that a "good enough" model is going to be fine.
These comments were also around for the Opus 4.5 release.
Prices for the same task (ARC-AGI-2) have since dropped by factor 10 if you compare to Opus 5.5 Low.
No it is not. Only maybe for the noobs or vibe coders.
People who aren't afraid of rolling their sleeves into any code base? The difference is practically zero.
I agree; yes, I can see that they need a bit less hand holding each cycle, but I also see these "frontier" agents do some absolutely dumb shit that I have to correct and then I'm wondering if I'm the looney one here.
Maybe it's because people stopped watching what their agents are doing and stopped looking at the quality of the output. But I still see agents being absolutely mindless like a junior dev.
Recent example: it updated an an API to add newly released models to the backend. There's a list of models that require specific configuration for the reasoning effort and temperature or the API call fails. GPT 6.1 Sol misses this and code fails at runtime because the newer models need to be added to the list for special handling of temp and reasoning. Fixes it for one model and tests it for that model using an E2E test. But doesn't test the other models that were added for the same error condition...I had to explicitly ask it to do so and it finds them and adds them to the list and says "that's on me."
Yeah, not that smart.
You're not looney at all. Frontier models do dumb things all the time, especially on mature codebases. Just yesterday Opus 5.5 butchered the OOP model in a codebase I work on - it duplicated a load of classes that should have just been subclasses. A junior checked it in very satisfied that it was perfect. The LLM review passed it, the tests were fine, and it implemented the feature successfully. It's just the code design had poor taste and poor long-term maintainability.
I keep seeing this kind of thing over and over, and honestly it's not got _that_ much better since the big breakthroughs about a year ago.
For sure I happily vibecode stuff without worrying about it when it's a greenfield project, and if the LLM has written it entirely from scratch then usually it's well structured and sane. But making changes in messy, mostly human-written mature codebases is still a minefield.
I’ve been saying that. When you have no idea what you’re doing, you *need* the latest greatest model because it’s the only way to reduce errors.
For people who have some expertise, the models accelerate the grunt work, but you’re the one validating it.
Try a bigger code base or more complex stuff and you will easily see that the solution, speed and amount of problems Opus5.5 solves vs older models is relevant.
I am frequently running agents on a multi-microservice application workspace where I really need the 1M context windows, because they are filled to the brim when implementing features that require changes on several services and APIs.
This works fine with Opus 5.5. But it also works fine with GPT 6.1 Sol, Kimi K3 and MiMo 2.6 Pro.
It doesn't work equally well with Sonnet 5.5, interestingly.
I agree that Opus and GPT are almsot surely better, but so many real users are nervous enough about giving Anthropic and OpenAI access to all of their internal information that they may be willing to stomach worse models if it gives them more security.
The real question is if this model is good enough that it can still accelerate work, and not be a hindrance to real work like older Mistral models often were.
If they can do that, they'll have customers.
The Navier-Stokes fiasco made me push for local/controlled models very hard. "Can't rule out" that they stole data (backed up by their backdoor offers of sharing credit).
If these companies will steal from deep pockets like Disney or Sony (some of the most infamously litigious copyright trolls to ever exist), they won't think twice of stealing every bit of code you upload to them.
If your code passes through an AI company's servers, you can assume you just gave it to them. In turn, when your competitor tries to copy that new feature you just added, the AI is now trained in exactly how to copy you and eliminate your competitive edge. Unlike your employees, the AI isn't bound by the same rules and even if it were and violated them, your company probably doesn't have enough money to prove it in court (and that's if we somehow reverse some of the stupid "AI is the most transformative use of copyright I've ever seen" judges who have drunk the coolaid).
Most companies could build the compute to run GLM or Kimi models for way less than the potential loss due to IP theft from using third-party systems.
I would also add to this, there are ways to use customer data to improve your model outside of just using it as “training-data”.
A simple loophole, use the code to create an RLVR environment where the resultant code is the end goal / max reward. Technically the customer data is never trained upon, but effectively you’re using it. Even better, use the code as a seed to generate synthetic data similar to it and use that synthetic data as rewards in an RLVR model.
Unless you can host the ChatGPT model on your own servers, which I know some enterprises are doing, I don’t think there’s any hope of protecting your data / competitive advantage from these frontier companies. Better to be paranoid, than be commodified by these companies.
tbh to me if the AI company writes all of your code & your eng don't even review it anymore then… the AI company _controls your company_. maybe that's ok if you make widgets but less ok if you do anything in dev tooling, security, or [insert market they may suddenly decide to compete in].
If I'm not mistaken, they did rule out that they stole data from the mathematicians in question.
IIRC they ruled out that humans knowingly stole that data, but they didn't rule out that the AI agent might have
I don't think that's correct.
> Following an investigation, we have confirmed that Buckmaster’s Codex prompts over the two months preceding this announcement and paper on September 8, 2026, could not have influenced the system in any way, including through training. The OpenAI internal model used for this result was developed through large-scale reinforcement learning on top of a previously pretrained model. Our proofs also differ significantly. In the Euler case, Alpöge and Buckmaster proved a result with external forcing, while OpenAI’s system proved a result without external forcing.
"We investigated ourselves and determined that we are not to blame"
Surely Sam Altman would not lie to anybody.
It's an improvement, but game changer might be a bit of a stretch. If I lost access to Anthropic or OpenAI models tomorrow, I would be annoyed, but would reach for a slightly inferior model. Last year I wouldnt be able to say the same, and rhe challenge is that the moat is drying up fast. Whether its general improvements in model training by other competitors, or straight up distillation of SOTA models, the moat is shrinking and the available capital and spend for American model providers is going to dry up quickly as competing good enough models are adopted by more consumers.
It's especially the case as more non-Americans look to self hosted models and domestic cloud inference providers using open models that the US providers who are still leading the charge need to drastically drop their prices and find a path to profitability in order to maintain their lead and retain the advantage they had as AI turns into a commodity (which is happening faster than I think even the frontier labs initially predicted).
Please, give it another 6 months and they catch up. The American labs are currently trying everything they can to block others instead of advancing their models, trying to build an artificial moat. The American models are not that great, they are good, and they have a lot of agentic workflows in the back, but its basically a hardware limitation at this point. Once the HW makers catch up, and we can move away from the Nvidia monopoly, things will speed up quite a lot IMO.
Just one more release cycle bro, I swear
We've been hearing the line about them only being a few months behind for a year now, during which time O/A have grown their revenue like 10x, haven't they?
We've also been hearing we're 6 months from AGI for about three years, and here we are.
"Now, here, you see, it takes all the running you can do, to keep in the same place. If you want to get somewhere else, you must run at least twice as fast as that!"
Phantom Tollbooth?
Nope, wrong bit of slightly-surreal children's literature. Through the Looking-Glass and what Alice Found There by C L Dodgson, otherwise known as Lewis Carroll.
How is that relevant to this discussion?
Those are two different things. The market is expanding, so even if competitors are catching up, you can have your own revenue, in absolute terms, grow.
That's fair, I left out a part of what I keep hearing in relation to the open weight models only being a few months behind, which is that they're going to eat the big labs' lunch, but then the big labs keep growing...
> but then the big labs keep growing...
Yeah, I could buy a bigger and bigger vehicle when a new one goes to market, but using a semi-truck to drive myself around would not only be significantly more wasteful but also more inconvenient and limiting compared to a sedan. Very much like advanced closed models vs performant open ones.
And yet the labs keep growing revenue like crazy…
A lot of big enterprise customers are already starting to question their spend and there is more and more pushback in regards to sharing private data with both AI orgs, even with the enterprise agreements in place. There might be growing revenue at the moment, but enterprise finance moves slowly, so the next couple of years will be the deciding factor.
The thing is O/A have been much louder on pacing the frontier, lately.
And yes, open weights are still behind, but are catching up.
I use Opus 5.5 at work.
I use MiMov2.6Pro, DeepSeekv4.1Flash, GLM5.3, Hy4, Qwen3.8 and KimiK3 at home. Opus5.5 is not a game changer.
What is your hardware running GLM5.3?
I do a broad amount of diverse experiments/projects I always wanted to do and throwing Opus5.5 against it just works
I have to admit, Sonnet got really good too.
But Opus just uses tools, a broad spectrum of it, etc. it feels like sure if you add some router behind it you could split it up if you need to but if you give me the choice, its opus allll day long.
People say this exact thing every single time a new frontier model comes out.
Let me know when the game changes are more than a month apart.
> X is such a game changer
I hear this literally every other week about whatever the newest FoTM model is.
Unless you can provide concrete examples of things you can do with them that you simply couldn't do with last week's model, it's absolutely meaningless.
Less and less work requires a frontier model though.
My todo app generator does not need opus 5.5
Yeah, I agree with this. I think the "the models are good enough" narrative is a myth. I've heard it so many times over the last year, but the model number keeps changing...
There is no ceiling on what you can accomplish with more intelligence, so there will always be a market for the best models, and that market is likely to just keep growing. If Opus 13.5 can one-shot a profitable company or discover a new disease treatment or whatever you can think of that a swarm of relentless super-geniuses could accomplish, companies (and governments) will throw money at it.
I also think there will always be a market for many sub-frontier models that will continue to grow rapidly as well, because "good enough" is definitely a thing for a given task.
One could still argue that models are good enough for a given task. I primarily use Opus at work for writing code and I realized that for my usage the intelligence of Opus 4.8 is more than enough. Sure the newer models are better but I can still do my work with having access to newer models
>If Opus 13.5 can one-shot a profitable company
No company would ever release such a thing
Despite Opus 5.5 got really bad the last days for me. Looks like they nerfed it again. This is extremely unreliable.
Or maybe they secretly believe you are trying to distill their models and are deliberately degrading your experience. Who knows with them?
You are laughing. Until it happens to you! :-)
It will become winner take all when AI companies manage to really get value from user logs.
Right now they don't even get good feedback from local sessions - I can see it make the same mistake two days running, and then months later when a new model comes out, presumably trained on my data, it still makes the same mistake.
law of diminishing returns? i.e. any reasonable frontier lab will have enough user logs...
My theory is there is actual knowledge in the user logs.
When a user says "I'm struggling to undo a bolt on my 1952 Mustang" and the AI responds "try hitting it with a hammer" and the user replies with "that worked thanks" - that is a tiny piece of knowledge which exists nowhere else.
Future AI's can say with more confidence that hitting it with a hammer will probably work.
Across billions of conversations, that can add up to more knowledge than all books.
I strongly disagree with this "early days" framing.
AI is an idea 60 years old. We are on the 3rd or 4th generation of AI development. Three years into the current iteration of products.
This is not early days by any measure. LLMs are a result of a very, very mature research field.
We haven't reached RSI yet. Once any entity reaches RSI, the runway scenario will happen.
Truly, this is what the Lord's prophets have revealed to us! (Eliezer 11:52) Keep strong in your P(singularity), for when the Kingdom arrives, He shall judge us in His righteous glory, whether to eternal annihilation, or rebirth and life in His Memory Eternal!
Assuming RSI is something that is possible as you envision it in the near term. I think that it will happen at some point, but I think we could still be a long way off. I don't think anyone can truthfully say that it is right around the corner.
What do you mean "good enough"? Did you mean "large enough"? ;)
Disclaimer: I'm not sure how much of an IYKYK factor applies to this joke.
Yes, so far the competitive dynamics feel more like cloud computing than web search.
this will be quite obsolete by the time the weights are released. still though, it's always good to see open weight releases
Especially with Mistral taking a fraction of the investment of the big guys. They can maintain the position pretty comfortably just by staying within a standard deviation of the leaders.
Hard to do that with a commodity that's easily replicated.
> Its quite interesting to see that at least the early days of AI so far have not been a winner-take-all runaway acceleration game where catchup is impossible.
Mistral is also an European company. As we live in a time where the US regime is engaged in pyrrhic geopolitical tactics, it's good to know that it can't threaten to cut access to models during s period where everyone is rushing to incorporate them more and more in our life.
I think it's a mistaken belief that AI as we found it is the exponential runaway train.
So it makes sense, since all you need is compute, that there's a ceiling and specialization is going to be more valuable then some super AGI.
Especially since the worst people seem to be the ones who think they'll all run away with the bag.
I mean, Mistral is about 9-12 months behind here when you look at its overall benchmarks versus the models released around a year ago.
Sounds ok to me. Claude was fine at the start of the year, and now with Mistral you also get EU sovereignty? I'll take that.
I don't think it is, and I think that is what will pop the bubble. All these companies have winner take all valuations, and that won't happen.
... unless they can legislate it, which is why they are flattering heads of state and scare mongering about dangerous AI.
On a purely technical level, maybe? But in terms of actual revenue, is there really any chance of anyone catching the big labs?
Obviously, this is only a valid question if you don't believe that open weights are about to eat their lunch and their revenue is about to collapse, or they're running a super unprofitable ponzi scheme propped up by investor money that's about to collapse like a house of cards. I don't find those positions credible at all though.
If you do, then this question isn't really for you, as I'm more interested in thoughts from those who think that OpenAI and Anthropic in particular are about to be the largest companies on earth in a couple years. Could anyone catch them at that point?
> But in terms of actual revenue, is there really any chance of anyone catching the big labs?
I don't know about revenue, but I suspect multiple other labs are already beating OpenAI/Anthropic on profitability. Staying on the frontier is expensive, and it's hard to recoup those R&D costs when you have a bunch of other labs nipping at your heels.
If you concede the previous point, then the only way for OpenAI/Anthropic to keep growing long term is to swallow the whole economy (i.e. mass job replacemnt), and that's a bet I wouldn't take.
I think the actual plan is to swallow a good portion of the job market. It’s the only thing that makes sense and I hear VC podcast debates on which percentage of jobs justifies the market cap.
Maybe. I can't freaking wait for the IPO filings so we can finally put all this to rest. (haha, like that'll actually put it to rest on HN, but at least we'll have better data)
The big lab revenue may not be catchable, but im not sure it needs to be.
If they can carve out a niche of industrial and governmental partners who rely on them for sovereignty reasons, it may be enough.
I completely agree, I think AI is a vast ecosystem will all kinds of profitable niches and sub-markets.
It's an open question as to whether or not superintelligence will create a monopoly/duopoly. My opinion is that it will.
They are very unprofitable…? I don’t think that’s really in dispute. We haven’t yet seen a profitable frontier lab and model pricing remains fairly subsidized
No, because compute, not model ability, is the moat.
The second moat is convenience, which all the big labs make it (comparatively) easy to glide into their models.
I don’t know if convenient is a moat when it makes switching very easy
How do you figure? I haven't met a single person who doesn't use Claude or Codex for programming in any serious way.
Then you dont know people working on highly sensitive info with stringent privancy concerns.
they just use claude on bedrock.
I'm happy I've never met you.
It's a retrain of asian model.
Seems silly not to have predicted that 10 years ago. I feel like it's long been obvious that smarter models being available will mean way easier cheap synthetic data and access to tools that will speed up competitors as well as consumers.
Man, a lot of this discussion sounds like people cheering for the last kid crossing the finish line.
Surely we want competition and Europe involved in that, but at this point I have grown used to either American labs smashing the frontier remarkably fast, or Chinese labs getting way, way closer than you would expect them to.
Mistral’s progress, regrettably, feels much slower. This model doesn’t knock anybody’s socks off. The model is (and I hate to be this harsh) mediocre, and this mediocrity has also arrived months late.
This is a pretty grim prognosis for European AI.
> Man, a lot of this discussion sounds like people cheering for the last kid crossing the finish line.
Sometimes it's ok to cheer for the last kid crossing the finish line because they're actually running a totally different race, and winning might look completely different.
When I look at what Mistral does vs other organizations I'm impressed:
https://isaiprofitable.com/
They aren't profitable yet, but they're a lot closer than most and they're doing a hell of a lot with very little.
Pointless racing story:
I was in high school track with a really tough guy who was just not a runner. We went to a pretty messed up high school and if you screwed around in track practice sometimes the coach would make you run a crap race at the next meet, like steeplechase or hurdles. Well this guy and a few others screwed up and coach made them all run hurdles at a meet.
He hooked every single one and fell on his face. Every time he got back up and kept on running. By the time he hit the finish line his knees were bleeding halfway down to his ankles. We cheered like hell and he was smiling ear to ear.
Coach quit punishing us with races after that.
Mistral is more profitable as a general percentage sure, but take their overall revenue: about $400 million annually.
Good, until you compare that to OpenAI’s behemoth $70 billion.
Sure OpenAI has much higher operating expenses, but which position would you rather be in? If OpenAI cut off all training and R&D tomorrow, would their 1 billion users simply vanish? Could Mistral ever realistically acquire 1 billion users?
Point being that profitable isn’t telling the whole story. And I don’t know if lack of resources is a great excuse, when Chinese labs are doing more with less, and it’s ultimately up to every competitor to win as best they can, regardless of their circumstances.
Great story you told, though. My old track coach was also an asshole, but of the more lovable variety.
I think right now it should definitely be a race to profitability, because revenue will not save you from unsustainable finances, no matter how large the values are.
I think OpenAI is currently far enough ahead that they could just cut training and R&D for a while (maybe even 1-2 years) and rake in some money before they'd need to start again, but if a smaller company finds a way to keep doing R&D while also being profitable, it doesn't matter if they're a little behind in quality, they will win in the long term.
> I think right now it should definitely be a race to profitability,
I haven't been listening too closely, but this is the first realistic mention I've seen that actually says what the "AI race" is or should be about.
>when Chinese labs are doing more with less
I need a source for this. AFAIK, the government heavily subsidize those labs.
I read that site quite differently from you. You seem to be analyzing absolute differences but ROI is really about ratio of spending to revenue.
It looks like Mistral is middle of the pack, behind Anthropic and ahead of OpenAI on that front. All of those labs are way "ahead" of the cloud providers, but those providers are building infrastructure, not just training models, so it's not apples to apples.
Yeah that's a good point. The numbers are smaller and I was looking at it absolutely.
Speaking from an absolute perspective I do think they are doing more with their money than either Anthropic or OpenAI.
Proportionally to income, Anthropic revenue is two thirds of its expenses, Mistral is a half, and OpenAI a little bit than a half. They aren't that different.
This is a great story. I don't think we should expect someone to win in the first races they participate. There are football teams that never win the championship and they always play and strive to be better. Mistral it's definitely improving with each release and props to them for releasing open models
> This is a pretty grim prognosis for European AI.
I think it's an incomplete read. What's the point in competing for a sizeable percentage of your funding when the finish line is incrementally being moved each month? Better spend it on leapfrogs which they seem to have done.
Meanwhile Mistral have a natural ace in their pocket with respect to regulation in the form of CADA and the Cloud Sovereignty Framework. I can't think of another company that would qualify as SOV-3 under that regime
So we’re supposed to cheer on companies that make worse products and only exist due to regulatory capture now?
Political polarization is turning the world insane.
As opposed to what? Companies that steal the collective intellect and intellectual property of the entire world? Yeah, what a great alternative. A world where all of your hard work is blatantly stolen.
Ah yes, a red herring, plus my favorite HN hypocrisy, all in one comment.
The 'information wants to be free' crowd complaining about protecting intellectual property now that it threatens their income stream...has to give the patent trolls and the RIAA a delicious bit of schadenfreude.
I invented this react button style with 8px border radius, ChatGPT stole it from me, I want rents!
People were extremely dismissive of chinese models until recently. They went from 1 year behind frontier to 6 months behind frontier to 3 months behind frontier extremely fast.
To be clear I'm not trying to dismiss European AI. I am a proponent of it.
But Chinese models have very much earned their place. The same cannot be said of Europe, so far.
Most comments you've made here sound like you're dismissing European AI though
I think one can be realistic about where things stand now, while also hoping for more in the future.
Serious people haven't been very dimissive of Chinese models since at least DeepSeek-R1 in January 2025. Sadly, we in Europe are much behind.
Important to point out that these 'X months behind frontier' really refer to the public frontier, and not the actual frontier, which private companies are free to protect indefinitely. Perhaps open models are in actuality 18 months behind the actual frontier - how would any of us know?
Because we have access to these models and use them maybe? Because of benchmarks? It does not feel like there is much moat, look you can even use small models these days and do productive work. It is rather that private ones starting to get into diminishing return area which without moat is quite risky.
People are cheering for a kid that is gaining ground in an ongoing race.
Hey I’m cheering too, but this kid is less like a kid and more like a multiple gold medal winning Olympian, and our once great track star is now on his last legs.
Europe has been at the heart of technological innovation for a long time. In recent years, this feels much less so.
This is a pretty mean spirited comment that fails to recognise the realities on the ground. The US is a technological juggernaut which Europe just can't compete with, because it's not a level playing field. The US has multiple moats in the form of financing, computing power, tech talent and one language.
This is very much a welcome development and we should celebrate it.
I don't think it will matter in a year or so. We are clearly topping out on useful intelligence for an increasing amount of tasks, as demonstrated by more and more models reaching the "useful" barrier.
This barrier is not going to start moving dramatically. It will simply be mostly satisfied for most work we do. Mistral is going to get there, soonish, long before the economy takes an entirely different shape (in so far that even happens).
There will be super human intelligence tasks, tasks truly constrained by intelligence for quite a while. Those will be few and far between, relatively speaking. Mistral will have plenty of opportunity to capture the other stuff, with a fraction of the resources required that it took the frontier labs to get there first.
It’s not that I necessarily think you’re wrong—-it actually makes perfect sense to me that eventually utility from frontier models will have diminishing returns—-but this has been a talking point for a long time.
The prediction that eventually more intelligence will stop being significantly more useful has been wrong many times before at this point. We should be prepared for that not to be the case.
I think it's reasonable to expect much diminishing returns for work that we have a good sense of. Compare it to a new laptop, phone or machine in a factory: There probably are better versions than the one you are using, but how much better are they and will it help you do better work? Is it even enough to be worth switching?
(I understand that the considerations are a bit different when ostensibly the switch is as simple as changing an API call, but that's not really true, when you have the AI integrated into a larger system and need to run evals and adjust the framework around it at some places at the very least)
Then, of course, there will be entirely new work that benefits from every bit of additional intelligence, but the stuff that is useful to 99% of the economy right now is probably capped on some relatively finite amount of that (with adaption/integration going to be the bulk of work)
Like I said, I don't disagree with the reason of your argument. Rather, I doubt because of a certain epistemological humility.
We can sit here and predict things all we want, but these sort of predictions don't have a great track record. The way things actually play out with AI advancement is often different from the way our perspectives reason they will play out.
It's better then to rely on previous experience and evidence, rather than using reason alone.
Curious, why do you say the model is mediocre? I haven't tried it, so I can't pass any judgement... I've learned to distrust benchmark rankings. Are benchmarks and Artificial Analysis the yardstick you're using?
> Curious, why do you say the model is mediocre? I haven't tried it, so I can't pass any judgement.
It's nowhere mediocre.
It's toes-to-toes with GLM-5.3 which is one of the best Open Weight model available (With Kimi K3) for general reasoning.
I just runned it on code reviews right now and it was able to catch some thread safety issue than DeepSeek-4.1 didn't. And DeepSeek-4.1 is by no means a bad model.
It's better than I expected for Mistral.
That said, it hardly seems like the innovative underdog some comments here make it out to be.
GLM-5.3 and Kimi K3 released 2 and 3 months ago respectively.
For all I know anyone could train a model of similar performance by just applying already published research to a new large training run.
> That said, it hardly seems like the innovative underdog some comments here make it out to be.
It's definitively not a revolution, but it is a major jump forward compared to Mistral Medium 3.5.
Worth to notice that they explicitly said they still refining the model, so we might expect something a little bit better at the end of the month
For something trained on a limited budget and 4000 GPUs, it is quite respectable.
I'm just wondering if it's really the result of impressive innovation, in line with comments saying that Mistral might in the future beat a bankrupt OpenAI due to being more profitable or efficient.
Or if it's more a situation where I could have Opus 5.5 go through recent papers from the Chinese labs to apply these to a new large training run, resulting in a model that performs similarly to Mistral Large 4.
I honestly don't know.
I would do this calculation as some measure of innovation / resources, then probably we can put this work in context a little bit better.
And if we could do innovation / red tape, I'm sure it'd auto win :)
Any lab can make the lack of resources argument. Regardless of that though, it’s every lab’s own responsibility to do well, and to do well despite circumstances.
Is that fair? Of course not! But who really expected for everything to be perfectly fair all the time anyways?
We can do more nuanced comparisons that take resources into account. That will help us see how efficient each lab is, but it doesn’t help the labs overall. Mistral shouldn’t be held to a lower standard simply because they got outcompeted.
Sure sure but if I want to answer the question "is model quality a function of resources or there's something irreplaceable in the US and Chinese engineers that can't be replicated in Europe?" it's a good metric. And if the European Union or an European VC is going to decide whether to invest in AI in Europe or not, they'd probably like to know.
Fair point.
Doesn't matter, it's excellent, and it's European with European inference, which solves the pains of all my clients trying to build data lakes and processes on top of it.
Nobody in the real world cares about minor benchmark differences in money losing coding agents.
And nobody in the real world is giving Altman or Musk their data.
Enterprise users are not the only users, and many, many people in the real world regularly hand their data over to Altman and Musk.
You don’t have to thrilled about this situation, but it is undoubtedly the reality.
I wish you were true - but I still meet a lot of people handling sensitive data and using free or cheap version of ChatGPT or Claude with their customer data.
I think we will see some horror stories come out with data leak in the next years.
Yes, it arrived months late. The waiting was too long I started losing faith. But when I saw Mistral’s series funding news, I knew they are on the right track. 3 billion EUR is just too significant
Neat. Wait 3 months for the landscape to change entirely.
LLM development is jumpy. It’s hard to extrapolate very far ahead.
I agree.
When Europe does surprise us, I will be the first to commend their progress. But until then, this is where we're at.
> This is a pretty grim prognosis for European AI.
Is this better or worse than no news from European AI?
I thought they wouldn't release anything at all, it's not dead yet :)
Those are comments from Europe. The US is waking up now and I expect them to be much harsher.
I really want them to win as that's our last horse in the AI race, but ~200 research-oriented devs out of 1800 employees? I believe they agree it's pretty doomed and have pivoted.
I expect the us to be sidelining all conversations in creative ways to mention anything at all about us ai companies
What? The US has been awake for almost 7 hours now
First few models will always be slow improving and worse. The way to improvement is working your way through a gajillion evals [1], finding bugs, gaps, and curating training data (this part involves human design as well as raw inference compute) to fix it. This is very time intensive and can't easily be "done once and then everyone has lesser work to do" since every model is different. Well, one way to accelerate it is to simply have more compute, which mostly openai and anthropic have[2].
This is mistrals first 1T-scale model and I expect the 4th or 5th generation to be close to the best for many purposes.
[1] These evals differ from the public ones like terminal-bench, are sometimes model-specific, need real, diverse usage to actually create, and are held secretly since quality of eval is the first driver behind the next step improvement of a model.
[2] It is not close. This model was trained on less than 4k GPUs, whereas astra used north of 100k GPUs.
Mistral is not that new a player though. How can we give them this much grace when other players like xAI have done more in even less time? I don't think coddling Mistral helps them.
And to the point of scale and training cluster, so what? Not only do Chinese labs have smaller clusters with less empowered GPUs, compute is Mistral's responsibility. You can't take away from other labs just because they fulfill that responsibility better.
xAI has a lot of compute. Deepseek also has a lot, not as much though. But this is changing with their new 160k huawei ascend datacenter in inner mongolia.
The lack of compute is not really attributable in that sense to mistral. First of all it needs general investor and government willingness, which is easier in a larger economy like the US or China.
Second, you need widespread usage of your paid inference service for two reasons: one it pays off your compute cost, and two it speeds up the improvement process.
The vast majority of deepseeks paid customers are within china itself (since openai and anthropic services are not reachable from china) which gives it a market. But for someone in france, there is no reason to use a structurally slower developing model from mistral compared to using one from openai...except when data guarantees are needed, hence the landing page focus on sovereignty. As far as the dual use aspect goes, a model like this is more than enough, so the government will be happy.
My point is that pretty much anybody can make the case that getting compute is hard. Regardless of that though, it’s every lab’s own responsibility to do well, and to do well despite circumstances.
Is that fair? Of course not! But it’s not useful to remark on parity, since nobody ever expected everything to be perfectly fair anyways.
We can make all the excuses we want. All it results in is holding Mistral to a lower standard. Does that result in more fair comparisons in regard to resources to output ratio? Maybe. But in my opinion, this sentiment is detrimental to European AI overall. We shouldn’t simply accept European labs as perennial losers, or normalize lower expectations. That doesn’t help anyone.
Reminds me of Gemini 3.5 Pro
hurr durr
The blog post https://mistral.ai/news/mistral-large-4/
It's a bit below DeepSeek 4.1 Flash at about twice the size. For a model that was supposed to come out a few months ago this is pretty good. Mistral catching up to the chinese open-weight models is great news. Excited to see how they will build on that!
They will RL it to be better.
What I'm looking for with this is of it hallucinates less that DeepSeek with its sparse attention, and can be used in better summarization and content generation in multiple languages.
DeepSeek sucks with languages other than English and China.
The price is really good with ML4.
I noticed they compared with the previous best deepseek model, 4 pro 0813 for the two coding benchmarks (DeepSWE and Terminal Bench) and not Deepseek 4.1 flash. A glaring omission really.
Very nice progress. Also I respect putting Kimi on those charts. Regardless of if they are beating the Pareto frontier (not now), model diversity is a good thing for humanity — I’m hopeful for the team to keep increasing their gains.
Open weight, European, competes with GLM-5.3 on cybersecurity. What's not to like?
It's French, duh. They'll probably surrender half of the company to the Germans after receiving a strongly (compound-)worded letter.
I'm literally English and I feel the need to defend the French here, they're among the most successful military powers in the world historically speaking. It's hardly fair to judge a thousand years of French military history by the outcome of one war where they didn't do well.
One imagines it may have been a joke
Yea :/
As far as your sense of humor goes, perhaps you can find a different niche than disrespecting the memory of the 100,000 French soldiers who died in the early weeks of WW2 precisely because none would surrender? That 90s revisionist trope to shame the French for not being neocon enough is beyond stale.
Europe needs a lot of these. Quickly. Way to go, Mistral! Keep 'em coming.
I’d rather have one European AI lab like Mistral, with the financial firepower and compute to pretrain its own large models, than 20 weak labs that fine-tune Chinese models.
AI needs big bucks, and Mistral is Europe’s Anthropic.
Absolutely not, the last thing the Europeans need is a monopoly that would turn into a bureaucratic moat. Competition is always good and new startups should get the chance to challenge the ones that cannot compete.
just curious, how do we know that le chonk isn't just a fine-tuned chinese model? and/or distilled from US models?
It's in the announcement: "ML4 was trained from scratch on 3,800 NVIDIA Grace Blackwell GPUs in Mistral’s own datacenters in Europe".
https://mistral.ai/news/mistral-large-4/
Once it's open weight people will be able to inspect and compare it's tokenizer, architecture etc and tell.
it's extremely unlikely that they re-use anything from a chinese model, that would be obvious quickly, what's more likely is using documents produced by a better model to create synthetic data.
It would not take them so long to train it. Their pace would be closer to the Chinese models.
Two is enough for foundation models, these guys and Aleph Alpha. The gap is closing.
> Europe needs a lot of these.
Europe needs profitable AI companies, not money pits.
The US has made sure that Mistral has a large market in the EU by temporarily preventing non citizens from accessing Fable.
A lot of European companies now want a model the US can't cut off, but also lack trust in Chinese models.
Some of these will self host Mistral but most will pay them by the token. It's not going to be a huge market or a huge margin within that market but probably it'll be enough.
This is the exact mentality that makes the EU fall behind. If you don't want to invest in something until it makes a profit, you don't get the benefits of being a pioneer.
Look at the size of our capital markets. That is a suicidal strategy. We dont have to mimic US.
There is plenty of capital, it is just not flowing.
Like Sam Altman, you seem convinced AI is a winner takes all kind off technology. Can you explain why?
> Europe needs profitable AI companies, not money pits.
AI is a strategic technology with obvious national security implications. EU should invest in its development whether it's currently profitable or not.
Ah, national security. The end-all argument all Americans fall for.
How did Claude and OpenAI make the USA safer?
there are no profitable AI companies at the moment ... This is the exact problem of European startups, trying to make them profitable from day 1 while American counterparts (and Chinese) keep bleeding money for years. Europe will never have a Tesla, a Google or an Amazon with that mindset.
Pithy, but only really applicable to Amazon. I'll give you some tips for future attacks.
I'm not actually sure Tesla is a great company from any perspective so should be easy to find another sick burn for them.
Google is going to be a little more difficult. Maybe say something nasty about advertising? Or go for the monopoly angle.
Google treats its customers the same way as Amazon treats its employees. Any sane person stays away from its services because a bot can close an account for no reason, and there will be no recourse. A total shit company, like some illegal IPTV provider: no accountability, and when the service disappears, you have nobody to turn to.
As if European companies are any better. Have you tried getting customer service from any European ISP for example? And companies still treat their workers like shit, it's just you can put up more easily with it since you can't get fired immediately.
Very true. Experienced it. Any appeal is like talking to a blackhole.
Name 1 AI company that is profitable, and no Meta and Google are not AI companies
Like the US!
…oh…wait…
You have to thank the US investors who funded Mistral from the very beginning.
Mistral would have gotten a tiny and measly "EU grant" and ASML would never have invested later had it not been for the US VCs.
Kudos to the US investors for being such selfless benevolent charities.
thank you for this and all the other great gifts the US and its companies have given to the world! we all love the us!
Forgot your /s
Ah yes. Let’s thank the might US for providing pesky Europeans capital a fistful of dollars. But maybe let’s do it after Americans thank for Russian, Arabian, Chinese, European capital and workforce. After all this is what “Made in the USA” means.
-5 on omniscience? https://artificialanalysis.ai/evaluations/omniscience
That's not particularly great.
That said I love that they don't seem to restrict cyber capabilities to any degree and even lean into it.
If the best model for cyber attacks is open for everyone to use it just makes us all safer I think. Of course you then also HAVE to use it or otherwise you're vulnerable, which is a great distribution play.
https://artificialanalysis.ai/models/mistral-large-4 for the main stats
Imo omniscience correlates better to how useful the model is in practice than the intelligence index. But you have to use both together of course.
Too late to edit, now including AA-omni Score [0] and 'by Domain' Software Engineering [1]. Also added comparison to the eyeballs-median of the 'top 10' models and then you see that indeed Mistral Large 4 Preview scores miserable in the AA-Omni indices.
[0] https://artificialanalysis.ai/evaluations/omniscience [1] https://artificialanalysis.ai/evaluations/omniscience?detail...I ran it on my trivia game Redactle which features a redacted Wiki article. Mistral Large 4 is not very good. It can sometimes solve a game with ~40 guesses whereas the top models like Gemini 3.8 Flash or Grok 4.7 can one shot most puzzles. My benchmark here aligns with AA Omniscience. I also have a version where the text is rewritten to detect over fitting to exact wiki text which changes the scores but not the leaderboard order.
https://redactle.net/llm-leaderboard?view=vital-500
Gemini models are summarizing wikipedia articles all day (when being used for Google's ai answer), can we draw some conclusion from this, did they train it extra well on wikipedia content?
>If the best model for cyber attacks is open for everyone to use it just makes us all safer
Issue is..
I don't believe for an instant that any of us, including US citizens, get access to the best models for cyber that the US has. I think any adversary would have to assume the models in use by the US side are unreleased.
US is not the only one dealing under the table by the way, I also think everyone should take China having unreleased models as an operating assumption at this point.
So Mistral is the best that the public gets access to. And that's if it's even the best? Benchmarks and pragmatic work have often been shown to be two radically different things in this industry.
If the best model for cyber attacks is open for everyone to use it just makes us all safer I think.
The NRA approach to AI safety.
This is awesome, one of the coolest Pokemon ever too for those that don't follow that universe :)
https://bulbapedia.bulbagarden.net/wiki/Lechonk_(Pok%C3%A9mo...
I wouldn’t mind this monster to be the mascot for mistral 4 large. It’s cute
Was wondering what Le Chonk meant (not French).
It's a joke, chonky is used to refer to fat cats, and there was a joke meme over the summer that Mistral are going to release a new model, Le Chaton Fat (chaton is kitten in French). The name is a nod to the memes.
Le Chaton Fat is still too dangerous to release.
Only model Pontifex considers a possibly conscious being. Obsessed with jumping in boxes with Cesium and some poison though, even escaped the litter box during testing, probably an alignment issue…
But it would be "Le Fat Chaton".
> Le Chaton Fat
I've read it a few times and I always processed it as "Le Chateau Fat" ("The fat castle"), which I guess also kinda works. I guess it is mostly because my french is so rusty I forgot (or maybe never learned?) what chaton meant.
The main thing I always get away from the comparison tables of these "big" models, is how well Deepseek v4.1 Flash performs. While still being the cheapest model by a long shot.
Beyond benchmarks, does it in day-to-day? Have always struggled to get competitive performance out of any Deepseek release going back to V3 vs Z.AI and Moonshot models. Maybe I really suck at whatever is needed to make DS models fly, but even tailoring my suite hasn’t gotten me far when I tried with V4 Pro. Happy for anyone who is able to leverage their models well, wish I’d be able to crack how to leverage them.
Will say their research is some of the best reads in the industry and I could not care less about their model release cadence as long as papers keep coming.
Yes. V4.1 flash performs really well. I don't know what to say, maybe you don't believe, but my team has been using it mainly for almost a month now for programming. It is as bad and annoying as any of the US sota, but costs pennies. And if you look close enough you find providers that can push it 300-500 tokens per second...
Maybe it is due to us being all very experienced devs. And can steer the model. But my daily routine is just to have 8-9 Zellij tabs open, DeepSeek in omp in each, and grind research and code day and night. Really nice model...
Hold up, 300-500tps at p50? At p99, even something like Opus 5.5 on fast reliably hits 330tps+, so that'd be expected but if truly p50, wow. And no regressions in tool call and structured output vs the official endpoint? Would love to test that with my evals, please share.
https://www.lithosai.com/
https://coralbricks.ai/
With structured tool calls you do want to be lenient on parsing, but that is a very small price to pay when the model is 100x cheaper.
Is GPT Luna 6 dethroning Deepseek V4.1 Flash? It's price seem to be undercutting flash at a relatively similar capability.
Yeah, very similar benchmarks at 1/4 the price. I've been pretty happy with Luna 6 though I still think the gap between small and frontier models is larger than many people want to admit.
It is not super good in long horizon tasks and worse than 5.6 in our evals. It really failed the agentic evals where DeepSeek, Kimi and Opus are the winners.
It is great on creating summaries and content.
I'm at a loss as to what to do now. I've been wanting to support Mistral for so long. I struggled on with Mistral Medium 3.5 for longer than I should have (although I also think it taught me some valuable process lessons).
Recently I switched to the Mistral hosted GLM-5.3, this worked very well and powered through a tonne of work. Unfortunately, I also completely maxed out two subscriptions within the space of six days this month. One can't stack subscriptions with Mistral, so I'd have to register a third account for another subscription, which will be annoying with changing API keys all the time. Sure I can switch to pay-as-you-go API, but that adds up really fast. The Mistral dashboard shows that a Vibe CLI monthly subscription for €18.44 actually provides €255 worth of API use (apparently, and I tried to check this with Support but it seems like they were intentionally vague).
After maxing out my Mistral subs this morning, I dropped $10 on Xiaomi to try MiMo-2.6. So far so good, seem to have done a lot of work for the $2.85 I've spent, and Xiaomi prices are still much better than the Mistral introductory offer for Le Chonk.
Not sure where to jump.
Edit: Not being able to stack subs is my biggest gripe with Mistral. I'd probably pay them $100 per month (5 subs worth), but I'm not going to switch to the pay-as-you-go API and burn much more money for the same amount of tokens. Instead, I've taken that extra money elsewhere. If they just allowed one to keep topping up subscriptions on the same account it'd be grand. Or even a bigger single subscription. Make a $100 tier with five times the capacity.
> Recently I switched to the Mistral hosted GLM-5.3
> [...] I also completely maxed out two subscriptions [...]
TIL that Mistral provides hosted GLM-5.3 and that it has a subscription model as well.
> Not being able to stack subs is my biggest gripe with Mistral.
If you were to use a gateway like GoModel[1] then you could use Virtual Models[2] to load balance across the subscriptions. For that to work, though, you would have to register the same provider as many times as subscriptions you have. It can all be defined in code, so you'd only have to do this once.
[1] https://github.com/ENTERPILOT/GoModel
[2] https://gomodel.enterpilot.io/docs/features/virtual-models#l...
EDIT: formatting
> Not being able to stack subs is my biggest gripe with Mistral.
Haven't tried it myself, but on Discord they've recently discussed this: Sign up for a team plan with multiple seats, only use one seat, usage limits are shared. Not against TOS.
You can confirm the allowed usage amounts with something like ccusage or tokscale
Yes, but I still don't know where the truth is.... The Mistral dashboard shows a drawn down on €255 worth of "credit" when you have a Pro subscription, which costs €18.44. There's no way I'm going to use the pay-as-you-go API if it means I'd be burning €500 in the next six days, as I just ostensibly have in the last 6.
I appreciate I can set a monthly spending limit for the pay-as-you-go API, but I'm just not willing to find out how far €100 will go, when I know it will go further elsewhere.
I just wish they had that €100 subscription tier, for the equivalent of €1k pay-as-you-go use.
I am on the same boat. 3.5 medium was disappointing, glm 5.2 and 5.3 were very welcome. But I still cannot leave anthropic or openai. This chonky model being expensive, mistral’s subscription quota will run out ridiculously fast
Seems hard to get customers at that price range when you're competing with open source models that are 1/2 - 1/3 the price but with similar capabilities.
People usually buy the cheapest, like Deepseek or GLM or they spend on Anthropic/OpenAI subs.. Are these models in the middle getting any users?
On a side note, I wonder if this was the popular free Space Bunny model that left openrouter yesterday.
If you are a EU company worried about your data then mistral is your only option. Think about EU military companies. They can't use US and Chinese models.
why couldn’t people just use chinese models rehosted in the EU? They’re open weights so anyone can serve them for any data residency requirement
> why couldn’t people just use chinese models rehosted in the EU?
Guess what Mistral themselves do?
If you want to use the model for something, where hiring a Chinese national to work on these tasks is a no go. Using a Chinese model is going to be a problem as well. Some tasks and industries are so sensitive, that countries will not allow you to risk, that the model is aligned with Chinese interests and not yours.
Because nothing forces the Chinese to make their next SOTA model open-weight.
I feel like that would be a big strategic mistake if you're European defense companies. What if the Chinese just stop releasing open models?
I think you misunderstand how data processing works in a LLM. You absolutely can download the weights of a chinese model and run it on hardware you control.
And Mistral also do exactly that. They host GLM-5.3.
This strategy only works as long as China continues publishing weights. Domestic training capabilities are one way of mitigating that risk.
Excited to hear this!
I barely use anything outside of cheap Chinese models on OpenRouter anymore. They are simply (more than) good enough for most of the things I do.
This model looks reasonably cheap. Though not deepseek levels.
Going to test it with Hermes, wondering where it will land in term of capability.
Bon chance, Mistral!
I've been dreaming of this for a simple reason: the french prose combined with GLM 5.3 reasoning capabilities.
GLM 5.3 is incredible because for the first time with an open-source model, it feels.. enough. I don't need much anymore, this model is great in everything. Except a thing : speaking french.
If the benchmarks are true, I'd be glad to switch entirely to Mistral.
Refreshing to see this.
The pricing ($.68 in/$.07 cached/$2.09 out) makes it much cheaper than Kimi K3, GLM 5.3, and Meta Muse Spark 1.3. That's great!
But also much more expensive than GLM 5.3-flash and Spark 1.3 Contributor (the Meta-takes-your-data pricing of Spark 1.3).
So, I think it would have to be significantly better than GLM 5.3-flash to be worth it. GLM 5.3-flash is already very good.
You are quoting the discount pricing. It is 50% off for the next two weeks. The blog post has the real pricing up front in the card on the right side: https://mistral.ai/news/mistral-large-4/
After that, it will be much closer to GLM 5.3, but you can also get 5.3 in their API! I dont see people really talking about that.
Good point.
I still need to evaluate it for my own workloads, but if you trust the benchmarks, seems about on par in quality vs GLM 5.3 Flash
Source https://artificialanalysis.ai/models/mistral-large-4?total-c...
That is the promotional price, after which it doubles :(
One step closer to Le Chaton Fat.
Can't wait until Mistral starts adopting fancy product line names.
Soon we'll have Mistral 6 Chaton, Mistral 6 Guépard, Mistral 6 Tigre, Mistral 6 Dents-de-sabre, Mistral 6 Beast King, etc.
Panther, Jaguar … Wait, wrong company.
I am waiting for Mountain Chaton to iron out the kinks.
Le Chaton Fat development was cancelled internally after it broke loose into the treat box.
Excited to try this. The low costs v. benchmarks alone here are worth a serious test. K3 has been my daily driver for a month or two now and it's dramatically reduced token spend (while not having much of a negative impact on productivity).
This was the era of the AI race I was waiting for.
Curious now that SOL is cheaper than k3 - is k3 still your primary workhorse?
Haven't tried it yet. It looks like sol is a hair more expensive on input, a hair less expensive on output ($2/in, $10/out per 1m, K3 = $0.82/in, $13/out per 1m).
Nice! Once they make it available through their API I will be happy to support them. My local Qwen3.8 27B is serving me well, but I miss the speed and concurrency that comes with subscriptions, and I am not currently paying for any.
Tais-toi et prends mon argent!
I believe it is already available in the API no?
Direct through them?
Yes: https://docs.mistral.ai/models/mistral-large-4-0
Yes and OpenRouter
How could I resist switching to a model named after my cat!?
Stats be damned irrelevant. The naming is good with this one!
you and 10k other redditors
Large 4?
> … Mistral Large 4. Unofficially ML4, very officially: le Chonk.
Curious that it scores higher than opus 5.5 in cybersecurity because the closed models refuse to comply. I wonder if that means it's more susceptible to offensive uses.
There was a section showing malware analysis which is great to have. I suspect everyone is using Chinese for this nowadays but now you can switch to this as well. Opus like GPT are restricted, it is hard to work with them doing offensive security and Fable is expensive.
The closed providers are serving useless, bricked models that are tainted with their shitty system prompts
Le Chaton Fat is here!
Le Chonk https://www.youtube.com/watch?v=hD51W2txi1Y
Yeah just saw that, I'm gonna keep converting it in my head.
Awesome!
All things considered I'm more inclained to pay for EU-based AI in the end (supporting local company and most likely being more aligned with EU regulations…)
I think it is important what they mentioned about using languages as well. Some smaller nations think they will culturally vanish because their “heritage” is close to nonexistent in the training data. If Mistral uses more of that language specific data it is a good thing and might help selling it to different non-english groups.
Maybe I'm missing something. Doesn't seem super impressive to me. A proprietary model with performance comparable to GPT-6 Luna and Deepseek 4.1 Flash, but at a higher price than either. The main selling point is that it's made in Europe... not very compelling, globally. I suppose maybe there is some niche where European-hosted open-weight models aren't enough to satisfy some EU regulation, where only the use of European-trained models is in compliance, but as a non-European I have no idea what that niche would be.
Side note: Wish this thread was more focused on talking about the model instead of debating about China and America. Whatever happened to staying on-topic?
You yourself also kind of pointed out why the discussion about USA and China is not off-topic. The niche Mistral wants to fill (afaik) is that it's European. I, as a European am really happy that they made such progress in so much worse (financial) conditions. I guess it's mostly geopolitics.
It's proprietary only until the end of the month, when its weights will be released.
I take that back, actually: I didn't realize they release them under a modified MIT license requiring large corporate users to enter into commercial agreements. So it is in fact proprietary. Sorry for spreading misinformation.
I created a brand new programming language to test LLM models on[2]. It looks like Mistral Large 4 fares about as well as Kimi K3 and Claude Sonnet 5.5[1]. Though it is much slower than Sonnet 5.5 and more expensive than both of those.
1 - https://bench.killswitch-lang.org
2 - https://killswitch-lang.org
Well, with this and Beam people are going to have to stop saying that western open models are dead. This is great news. Anthropic and OpenAI may have a bit more knowledge, talent and compute but they don't have a monopoly.
Investors looking for them to make monopoly profits are going to be disappointed The premium they can extract from consumers for their models will be capped. Tokens are likely to remain close to the cost of compute, a cost which is high but falling fast.
Also, yay Europe! Although the comparison between this and mimo 2.6 is not flattering...
I don't know why, but I personally find Mistral's marketing strategy much more appealing than that of other companies.
For example, there's something about Anthropic's picked design and their little Claude avatars that's unsettling to me.
Sans context, I really like the Anthropic and Claude's faux-academic, minimal-ish, intellectual-ish branding.
(Especially the way it looked ~12 months ago -- it's gotten more cluttered since then. Perhaps unavoidably, as the breadth of their offerings has grown)
But over time it's begun to feel like unsettling cognitive dissonance as their ambitions grow and the stuff to worry about has piled up.
> faux-academic, minimal-ish, intellectual-ish branding
By this you sourely don't mean the messages shown in Claude Code, where Pi would show "Working..."
Academic style: Cooking... Sautéeing... Julienning... and similar annoying faux-brogrammer moody status messages.
It's straight from The Sims [1], which seems quite appropriate :)
[1]: https://en.wikipedia.org/wiki/The_Sims
You didn't have to look at "reticulating splines" all day long though. Anyway that stuff came from Sim City.
The logo isnt a stylized butthole. That sure helps endear me to them.
I know the whole “AI logos look like buttholes” thing is a joke for most people, but I really speaks to the pornification of our society. I would never have thought “butthole” looking at any of their logos.
It reminds me of looking at my cat’s hind end, which is not a good brand association. Has nothing to do with “pontification”.
I didn't know this was a thing and I typed "AI logos look like buttholes" into Google and the AI overview response started "You are completely right."
Have Google's models achieved the ability to throw shade?
Good point, it could be any kind of sphincter.
If it looks like that you, you will be seeing it everywhere.
https://velvetshark.com/ai-company-logos-that-look-like-butt...
Some more than others
Why was I about to write this exact same comment, and why did you beat me to it?
You are assuming that OP is not a fan of buttholes
The entire Anthropic branding is religious kitsch - deeply off putting, but apparently quite reflective of their reality.
I think there's such a thing as throwing too many marketing and sales people and too much polishing and "refinement" at something. I see in new product announcements from Microsoft as well. It's like seeing someone try too hard to impress you.
They've polished all personality out of their companies.
And their cookie banner. Never thought I would like a cookie banner
I do my best to avoid any AI marketing because I extremely despise it. I just haven't quite decided on my new profession, yet, but it's either going to be with plants or with animals.
Mistral Large 4: 1050B, 49 Active
GLM-5.3: 753B, 40 Active
I was hoping for something that hinted at smaller models too, but I guess not.
Any competition is still good, especially now that the USA AI labs are starting to do regulatory capture.
Give them time. ML 4.0 was just pretrained. Mistral will certainly use it as the base for distillation and RL for smaller, better, more efficient iterations, just as the competition does.
It's competitive!
Good enough to show competence, and instill confidence in the team/company. Later releases can be more efficient.
I think it's a great release with that framing.
It's Mistral Large, they usually publish Medium and Small later
just keep RL frying it should get better...
Pretty interesting but I wonder apart from sovereign use-cases in EU, who uses Mistral and for what? How good are they compared to frontier open models like Kimi?
I agree sovereign markets themselves can be huge but just curious.
Off Topic - The Mistral website - Really nice design. My guess, built by a human.
The key factor isn't whether the author used an LLM, but whether they had taste and attention to quality and iterated accordingly.
My guess would be designed by a skilled human with help from AI and built with AI by a skilled programmer.
I am not sure most people will be able to run this locally, and compared to GPT-6 luna it is much slower, verbose and also performse worse.
GPT-6 Luna vs Mistral Large 4: https://aibenchy.com/compare/mistralai-mistral-large-4-0-hig...
What is AI benchy and why does it rate everything on passing 23 tests? Putting Gemini 3.6 Flash on top of the leaderboard doesn't pass the smell test.
My own tests/benchmarks.
There is no model passing all the tests. Gemini is on top because questions also include general knowledge, trivia or domain-specific questions, which Gemini models are way above any other model.
Even if models have similar scores at the top, it's useful to compare costs/response times.
Everything in the top 25 is above a 9.4 out of 10, so it seems a bit saturated? Also seeing that Claude Opus 5.5 is at #23 makes me think this is not the benchmark for me
Yeah, it is saturated.
It's actually quite hard to make a single task/question with few steps max that SOTA models fail on. Gemini are on top, because asking domain specific questions still breaks most top models, but not Gemini ones.
Claude is lower, and also most Antrophic models, because (historically) they have one of the highest refusal rates and also struggled more with precisely following instructions (i.e. outputting in a specific format/way).
You can assume that all models above a score of 9 are "good", then it's more a matter of preference for their response times, costs, personality and strength in specific areas.
Opus 5.5 are best at coding, and even art generation, but struggle a lot with built-in knowledge, when web search is not available.
From Guillaume
> The RL run behind this preview is still in flight and shows no sign of saturation -- we will release a final version before the end of the month along with the weights of the model.
https://x.com/GuillaumeLample/status/2107461898127954001
https://xxcancel.com/GuillaumeLample/status/2107461898127954...
Something looks off in artificial analysis. Benchmarks aren’t everything, but not even close to the Pareto https://artificialanalysis.ai/models/mistral-large-4?cost=in...
I guess lots of token usage.
Claims to be on par with GLM 5.3 in DeepSWE (from https://thenextweb.com/news/mistral-releases-large-4-a-1-tri...)
Does it have the ability to capture the market like OpenAi or Anthropic ? My point is, Regular/Average users of AI do not really care about benchmarking. Marketing decides which company makes it to the phones or PCs.
> Marketing decides which company makes it to the phones or PCs.
Right now a big part of LLM market is people using it for professional software development. Most of these users probably care about the quality of the model and also notice it during daily work.
For normal consumers, shure it doesn't matter. In the end the ai summary of google will probably be the most used as they are already exposed to it anyways.
Do we have any published data on the proportion of LLM market by the category of usage?
-software development -business productivity -consumer chat -other?
I’m sure the big AI vendors know.
We're probably fast approaching the scenario where the cheapest models will win out.
Benchmarks are better than expected! And probably got there without distillation ;)
Is there a reason to believe why they wouldn't distill locally running open weights Chinese models?
'AI sovereignty' is a nice sales pitch in the EU market.
It's more than a sales pitch. Given the current geopolitical climate if I were int the eu I would not trust any us tech company with anything. It's possibly one of the most important things they can do to survive the next 5 years.
Looks like they are doing 50% off to stay price competitive with DS Flash V4.1
Still second most expensive open-weight model. I don't care about cybersecurity index. And still can't beat Chinese models but good to see European in the game.
Is that a reference to LeChuck in Monkey Island? Love that game!
No, it's a play on the Twitter/Reddit memeing about "Le Chaton Fat", see: https://www.reddit.com/r/MistralAI/comments/1u6f0dm/what_is_...
You'd be surprised how much of the AI world is fueled by memes.
litterally, that's what an LLM is doing?
I think it's actually a wink and a nod towards the social media meme of "Le Chaton Fat," a fictional model that is jokingly attributed to Mistral, usually with century-defining benchmarks and unfathomable size.
https://x.com/i/trending/2066326562623127678
I tried it a bit and I like it! It is very fast via openrouter (significantly better than Kimi K3) on webui. Very verbose and starts to forget instructions after awhile it seems, but it gave me quite a lot of good info during a half an hour chat on C and embedded programming. I think I will keep urimg this as my main assistant for few weeks.
Excited to test this on my benchmark[1], but I'm guessing it won't fare better than MiMo v2.6 Pro which is currently the best open-weight model as far as I'm concerned.
Why isn't this on openrouter yet? Is there a better router that gets models much quicker?
1 - https://bench.killswitch-lang.org/
https://openrouter.ai/mistralai/mistral-large-4-0
I tried Mistral's last 3 large models, devstral and mistrallarge3 and the numbers were not even benchmaxxed, but just false. the models were so weak and garbage. Let's hope they are telling the truth this time around, we need more alternatives. There mistral-small and original MistralLarge models were awesome, hoping they are back!
Unfortunately Mistral Large 4 is benchmaxxed.
It also has guardrails, albeit very weak (which you can read in thinking traces).
I've had seen subagents fail with "hmm, hmm - hmm, hmm (...)" multiple times, it keeps trying to cheat at the task (trying to remember hallucinations).
GLM 5.3 (which Mistral also serves at 3x the speed!) wipes the floor with it
The terminalbench 4.0 score is encouraging as a sign of it not doing anything "stupid" when put in a proper harness.
Good to see Europe is at least a little bit still in the game.
https://docs.mistral.ai/inference/model-selection-guide?mode...
Cost is stated at half the price of GLM-5.3, which is quite interesting.
Seems to have problems with tool calling? When doing heavy research it makes a 'json db' but then has trouble working with jq.
I tried a quick abbreviated kimbench on their playground before bothering to do the whole thing.
Maybe I didn't really select mistral 4? Either way, failed completely on question 1 and the next 2 questions were completely off base too. I didn't bother to finish.
Not suitable for my purposes I don't think.
https://venturebeat.com/technology/mistral-debuts-large-4-le...
It's quite good in legal research - it's so annoying I can't use it within legal information systems.
At the end of the blog post we get this nugget.
> The pace of progress from here will be fast. Stay tuned.
A month ago it seemed most of the reviewers have given up on Mistral releasing a decent models, looks like the table has turned
What experience have people had with Mistral models for cyber research? are they as constrainted as Anthropic models? I use claude day2day but have the need for a model with fewer guardrails to pentest our own APIs.
They explicitly lean in to cyber work, and they appear to be very permissive from their marketing:
> This is particularly important in cybersecurity, where provider-level refusals can block legitimate vulnerability research and incident response, and where losing access to a capability mid-incident can itself become a critical security risk. ML4 pairs top-tier cyber performance with open weights and self-deployment, giving organizations both the capability and the autonomy to run advanced security work under their own policies.
> That top score reflects a practical advantage. Several leading closed models, including Claude Opus 5.5 and GPT-6 Astra, score near zero on the same test because they refuse to perform the task. Yet defending software often starts with proving that a flaw is real, exactly the kind of work safety filters in closed models can block. This matters even more as threat actors increasingly jailbreak those same models to support offensive cyber activity
Excited to see this! Nice that they are saying this is just a first step.
Give them more compute!
le chaton fat is real, my life is complete. Benches look crazy good for 1T.
Every few months, Mistral becomes active and drops a bombshell.
Is the reason for the massive gains in certain benchmarks due to distillation from the other lead models hence the slightly "under" pattern seen in the comparison charts?
Title is missing the official model name (Le chonk)
I tried testing it, but reasoning effort parameter doesn't seem to be there and output is sort of broken because it reasons directly in the output tokens...
Seems like they've finally made a genuinely competitive model since the original LLM craze, congrats to them! Glad to see some diversification in open weights providers.
Awesome!
All things considered I'm more inclained to pay for EU-based AI in the end (supporting local company and most likely being more aligned with EU regulations…)
Has anyone tried Mistral for the coding task? How do you find it compare to Claude, Codex or the Chinese models?
They have published their benchmark. Keen to try it too.
This sure is nice, but I've had less than satisfactory results with GLM5.3. I'd like Mistral to compete with Qwen3.8-Flash-Next a 120B class model that IMO is the first model that I can use for serious coding while running it locally.
I estimate it's coding ability on par with opus 4.6 (but opus definitely beats it on factual knowledge) Still it's a genuinely useful model, when everything else except Anthropic's models (and for only 3 weeks after it came out Google's Gemini 3 pro, before it got merged) are not to me.
I'd live to have one like that but EU made.
no hugging face link :( ... but hey its on openrouter yay
Here are my results
https://dach.peerbench.ai/compare?models=mistralai%2Fmistral...
Looks like a bit better than the recent Kolibri-1 but still below Qwen3.8 27B
We actually got Le Chaton Fat before GTA 6
Since the Chinese companies publish their research it would have been odd if Mistral didn't start catching up.
It certainly has helped OpenAI and Anthropic get their KV cache costs under control.
It's no secret that everyone is dis-stealing from everyone else.
I don't see how distillation relates to using the published techniques developed by Deepseek, Moonshot, Zhipu, etc
touche
if it's not available yet why have a 'try it today' header at all?
> "Try it today" > > There is still more to come. As we work toward releasing the weights, we will share further details on the model architecture, additional benchmarks, and our post-training methodology.
You can try it on their website, https://console.mistral.ai/playground
It's just that the open-weights aren't yet available (although the long delay is slightly annoying).
What's up with the name? It reminds me of my teenage self trying to speak in funny memes.
Without exaggeration, given a choice between models, I would pay for Mistral's model over Anthropic's based on the name alone, completely ignoring features or other technical considerations. The name is playful and is such a refreshing contrast to Anthropic's (and OpenAI's) doomsaying, scaremongering, and god-posturing.
Why don't they adapt the deepseek v4.1flash model instead?
> Trained from scratch
How are they training without pirating the Z library corpus and all that?
I interpret "scratch" to mean brand new weights. Not that they aren't training on a corpus of human text
Right but how did they get a corpus, how do they compete legally without distillation?
Glad to see progress, despite the ever-increasing sabotage by the EU bureaucrats
sorting the charts like that gives off weird vibes
https://mistral.ai/news/mistral-large-4/
sorting the chart like what? You just linked to the main page.
If you scroll down, most of the charts on that page are sorted s.t. Mistral's bar is right next to the worst competitor model, while the best competitor model's bar is positioned on the opposite side.
If one were to be cynical one could say that it's intentionally making Mistral's result look better than it actually is by making it harder to compare the bar heights.
So about 2 or 3 generations behind, just like they were a year ago?
I really hate this open weight but closed science approach. These companies just take from academics and all the Chinese companies that are doing good science, but without understanding the recipe it makes it hard to know where the failure points will be until your agent accidentally commits a crime.
Mistral doesn't publish the science.
Brother I think No one publishes the science
I love the name! Teasing the ones making fun of them.
Is there consensus on if this was https://openrouter.ai/stealth/space-bunny-alpha ?
Space Bunny Alpha is probably MiniMax M3.1 (rumors on Twitter since it seems to have a similar tokenizer).
Yes broad consensus had developed in the ten minutes between announcement and you asking if consensus had developed, and I'm excited to report that it consensed in the affirmative -- it IS Space Bunny Alpha!
sorting the charts like that gives off weird vibes
Had the same thought - feels chart crime adjacent
Bar charts should start at zero. If they don't start at zero, there should be a clear visual indicator that the chart has been trimmed without having to read the axis labels. I hate that this has to be repeated so often that it has become a cliché.
Number one in Sovereign AI. Join our Discord.
Previous one is barely in top 50 on arena.ai
I'm glad they're keeping at it!
I thought lechonk motto was just a meme!
Half price on open router right now
A bit disappointing to see it still lagging behind Chinese open models. Those Chinese models are pushing proprietary models to raise the bar, but we need equally strong non-Chinese open models to challenge the Chinese ones in turn.
Finally, a model small enough to self-host on my 2012 MacBook Air if I don't mind my house reaching room temperature in 2 seconds.
You might be missing three 0s on the parameter count, or what am I missing?
I'm assuming you need somewhere 0.5 to 1TB of RAM for the weights only
not to ignore you but how is it possible that I have zero karma
Someone downvoted you. Now you have -1.
Why is distillation weird?
at 200M tokens for the full AI suite run its not token efficient at all
I mean no disrespect but these are terrible numbers or am I missing something? It seems like Mistral continues to only be relevant for people that want a model trained in Europe. Too bad.
On Prem. Thats a bid deal for some enterprises.
also the benchmarks are not necessarily indicative of how well the model will perform in its own harness with its own skills.
Any open weights model is "on prem".
I mean usually the benchmarks make any model feel better than how they actually perform.
better than K3 and DS4, cool
Not bad only two major releases behind top tier. Edit : checked its rather 3 generations behind . Oh well
Where can it be tested?
Looks like it's about a year behind still. i.e. its intelligence is behind models from roughly a year ago.
https://www.vals.ai/benchmarks/vals_index
where does sit on the pareto distribution compered to Le Chaton Fat?
dont take it personally, i just dont understand why to release a model that is not showing new strong capabilities, why would anybody use this model and not Claude Opus.
Because screw Anthropic and OpenAI, that's why.
Amazing! just tested
impressive release this time by Mistral. bullish.
>> Unofficially ML4, very officially: le Chonk
Honestly just nice to see a leader in this space not take themselves so seriously.
europe finally getting into the race here.
wowza le models a heckin chonker
1T parameters -- ugh, open models keep getting bigger and bigger! Running them at home is getting ever more unattainable, especially for those of us with bandwidth-poor hardware like Apple silicon -- please continue releasing smaller models, too!
Massive fumble not to call it “le chaton fat”.
Too bad this got marked as a dupe, as it actually has benchmark info unlike the other page which is just docs.
The weird thing is how worse they are at things like coding than other open weights. You'd expect them to at least distill coding from other open weights to match them.
Pretty impressive. I genuinely wonder how Mistral hires talent when their salaries are so terrible. Guess there aren't many better places to work in Europe.
YES finally
Awh I was half expecting a zombie pirate..
Anyone have any indication when I can get my hands on a developer plan for this?
They do sell subscriptions I think
Don't believe Mistral. They're wrong. It's really called "Le chaton fat".
Also 1T-A49B. Weights currently closed but promise to open source them by the end of the month.
Great release movie.
>Don't believe Mistral. They're wrong. It's really called "Le chaton fat".
OpenAI's therapist: Le Chaton Fat isn't real and cannot hurt you
Le Chaton Fat:
Woah, this seems like a big deal (assuming the benchmarks are as good as claimed)?
Mistral slightly proving me wrong (and I'm not mad).
Now THAT'S how you name a model. Take note, others.
quick question why put GLM 5.3 at 61 while a quick check on DeepSWE 1.1 puts it at 69?
also they forgot muse spark at 75% while claiming they were outshining all US models?
>Frontier performance
* proceeds to not compare to Opus 5.5
They say "open weights frontier performance". Of course they aren't comparing to Anthropic, because they are not putting their weights online.
I'm talking about the section of the article labeled "Frontier Performance", which does not specify that.
Can we consolidate the posts? Currently there's 3 on the front page, basically all pointing to Mistral's messaging in different places.
Mistral wont win the AI race because of the model names. I wont bother an arrogant Parisian hipster with my insecure prompts who then plays with his moustache and responds with a judgmental "pfff"
Still a long way behind closed models sadly :-( The gap between open and closed seems the biggest it's been for a year or two.
I try to stay up on AI models, but I've given up on Mistral. I have tried using it too many times and it's the worst out of major models. Even Kimi and DeepSeek are miles ahead.
Seems like it's another European company that is only alive through government financing.
I understand wanting a home grown industry, but with AI models, just abliterate a DeepSeek model.
Total misunderstanding of the industry. Europe needs hardware, not a model that needs to be replaced monthly.
We need to maintain the capacity to build reasonable models in the meantime, until we have hardware though.
Remember also that this model is still in preview and won't finish post-training until in a month or so.
> Total misunderstanding of the industry. Europe needs hardware
With all due respect, I have no idea what you are trying to say with this comment and have no clue how "hardware" would turn Europe's tides. You explained nothing.
Do they need training hardware? Inference hardware? ASICs or GPGPUs? Edge hardware? Agent hosts? Faster cores, or wider ones? Taller systems, or more parallel ones?
Your vagueness completely undermines the authority that your criticism relies on.