This might be the first time I’ve favourited a post without upvoting it.
This is really sus. I’d love a more detailed explanation and “take” from AA on what problem they felt needed correcting and why things have shuffled the way they have. And how they can claim their index has relevance if positions can be switched around so easily.
It is embarrassing how much they suck up to american labs while those models only perform the first few days after release. With open weights you can be sure the quality stays the same unlike the closed source crap. How about a retest after some time?
The natural language choice in multilingual index seems pretty strange and biased to me, they tried to encompass a broad swath of Asian languages, but then there's like four European romance languages. They didn't include any Slavic one i.e Russian, or Nordic one like Swedish, or Finno-Ugric. And these romance ones are probably overrepresented in the training data anyways
I think it's pretty evident that Artificial Analysis loves to change the benchmark to place US models on the top, it's no the first time either.
The current placement of many models simply doesn't makes sense, GLM 5.3 is so much better than Astra at backend tasks, Astra does the most weird nonsensical bugs.
This is what happens when the benchmark team is funded by VC capital.
Even and especially for frontend UI, GLM 5.3 is a whole world better than Astra.
I can see (and believe) Astra beating GLM 5.3 on verifiable hard domains like maths, etc; but we’re at this stage where intelligence scores and even ‘frontier real world unsaturated benchmarks’ really doesn’t reflect real world use anymore.
Yeah, this smells. Ain't no way Terra went up and Deepseek so far low. I have had the later beat the former on multiple criteria by simply being less distracted.
There was a thread here about families of models responding in the same patterns to specific queries like "say a random word", and it takes less than two minutes to verify it is true. It'll take a bit of effort to not convince me this change wasn't reverse-benchmaxxed to screw everyone other than OpenAI. I'll be less distrustful after everyone is done IPO-ing.
Neither is the primary cause, I suspect. If you look at the coding index alone, you see more plausible results from a software engineering perspective (in particular, Terra no longer outperforming Fable) [1].
That especially the small flash models that are primarily optimized for coding tasks would perform worse at non-coding tasks is not per se surprising to me. The introduction of Terminal-Bench Science 0.1 in particular hurts their aggregate scores [2].
The general question remains of how meaningful aggregating the results of very heterogeneous tasks into a single numeric score is. (Aside from general questions about individual benchmark validity.)
Note: the cyber index is also interesting. While refusals make it somewhat opaque, even if you optimistically count all refusals for Opus and Fable as successes, the results are ... fascinating [3].
I'll add that I really want to know how Opus 4.x and Codex 5.x perform on the new index. Once we have that information, this is going to give us a better understanding of how meaningful the scores are.
Tinfoil hat time: but the way they waited a couple weeks to evaluate Mimo 2.6 Flash, allowing GPT-6 Luna to sit on the Pareto frontier in the meantime, really did make me think they might have some bias going on... (Luna was displaced when they finally evaluated Mimo 2.6 Flash, which had been released the day before Luna)
edit to add: there totally is benchmaxxing going on, but I'd be really surprised if it wasn't across the board. After all, how can any lab tell if their model has improved over a previous iteration, if not by comparing benchmark performance?
I’ve very mixed feelings on Mimo. The new models do surprisingly well on some tasks, but they don’t allow you to set the thinking level and the model often times spends next to zero time thinking about a problem. If you coax it just a little bit, you can get it to solve it, but this isn’t an issue with other models. Also it’s incredibly easy to get it to get caught in a thinking loop.
It is hard to tell what people are trying to say when they use undefined words with no particular meaning. This doesn't seem to change the relative position of the Chinese models much. The Chinese are a month or few behind and the state of intelligence is so volatile that the top 10 will be completely different in 6 months. I'd expect it to still be a mix of US and Chinese models with the US models leading.
Benchmaxing is defined as targeted post-training attempts to optimise for higher benchmarks. This doesn’t change the aggregate intelligence or capability of the model, so it can make a model appear more capable than it is.
Glancing over the methodology[1] made me curious about three things:
1. Is there a benchmark measuring how much better an LLM performs when you tell it "you are an X"?
2. Does telling an AI agent "you are an AI agent" actually help? Is it the LLM equivalent of reminding a senior developer they're a senior developer, making them feel guilty about being lazy and finally getting them to do their job?
3. What happens when the persona conflicts with the task? If you tell it "you are a PM" and then ask it to write code, does it struggle? Produce worse code? Spawn a sub-agent and tell it they're a developer?
I find it fascinating that some of the main areas people are interested in are using benchmarks with quite few questions: ~200 for coding and ~150 for science.
As long as that's the case, it's no surprise that there is often a perceived difference of quality between benchmarks and real-world use.
I know that running the full composite benchmark set is probably already quite costly, and also that the benchmarks have to be rotated regularly, but it feels like a significant untackled problem.
Or, are we witnessing reverse-benchmaxxing? Meaning that first some entities let players optimise for a given benchmark, then they update the measurement methodology so as to retain some players on the top ranks, while dropping others. As former Italian prime minister, and highly controversial public figure, Giulio Andreotti once remarked "Thinking the worst is a sin, but you’re often right."
I keep seeing that MiMo is in the frontier in terms of performance and cost but I see very few talking about it and the only ones that do talk about it are doing it in context of AA
It's also dog slow at AA index tasks compared to other models in its class. Makes me think they optimised for getting on the intelligence/cost frontier at the cost of real world usability.
I used MiMo for a couple of days waiting for my Codex sub to reset. I really liked it, its communication style was a breath of fresh air compared to how stale gpt-sol is.
chinese labs are probably benchmaxxing but i'd like to see when the index changes relative to the position of the best chinese models (last suspiciously-timed update was when glm-5.3-flash was looking really strong)
This seems to be a hit and a miss, there's no visual indication that opus 5.5 is better even though in 9 tests out of 10 it will outperform astra especially when it comes to niche domains and will absolutely destroy them in anything requiring raw intelligence rather than a naive search, naive search is what makes openai models extremely capable in scientific fields though.
The score hit on smaller models is a welcome change especially when compared to models that nearly 10 times larger.
Grok being punished quite a lot is also a very welcome change as the model is frankly not that good.
I would love to see few more points knocked off haiku by introducing steerability/creativity benchmarks.
Honestly it does somewhat match what I’ve felt playing a lot with all OpenAI / Claude / deepseek models
One thing I’m still surprised / unsure on is haiku 5.5 max besting latest deepseek - but also I think the last haiku was so incredibly bad I still hold some bias whilst using the new one.
(EDIT: A child comment has refuted the AA score comments, AA runs using their own harness.)
One of the new benchmarks is FrontierCode 2.0 by Cognition. AA seems to have accepted Cognition's benchmark scores without questioning, but they should have looked more closely.
On this benchmark, Anthropic's models are run using Claude Code; OpenAI's using Codex, and Gemini is run using... Cognition's own Devin (not Antigravity). Thus its score is much lower than it should be, knocking ~3 points off its AA score. All open-source models are also inexplicably run using Devin, damaging their scores as well.
This may even be intentional, if Cognition has beef with Google, since Antigravity was built using Windsurf's technology. But aside from that, it is to their advantage to make Anthropic/OAI look like leaders in the AI market. They are a company known to lie and exaggerate again and again, so I have trouble giving them benefit of the doubt.
It's interesting that the Stirrup scores are somewhat close to the specialized harness scores. Either Stirrup is quite excellent or is it that harness doesn't matter so much after a point?
Yes, that is true, you might be right. But I think such, because the moat in their wrapper layer feels very thin and does not make sense in a world full of 20+ wrappers around similar models. It feels better to "only" have to deal with Ant/OAI (and other pure wrapper cos). Competing vs. Google, Meta, and every Chinese lab on earth, every one using their own bespoke wrappers, is much scarier.
Additionally: competitive frontier = more commodified = high valuations under more scrutiny.
This might be the first time I’ve favourited a post without upvoting it.
This is really sus. I’d love a more detailed explanation and “take” from AA on what problem they felt needed correcting and why things have shuffled the way they have. And how they can claim their index has relevance if positions can be switched around so easily.
From my rough recollection, all 3 of the past updates all had the effects of decreasing the score of open weight models.
One was rushed and same-day updated as GLM 5.3 (iirc) got added; of course its main change to rankings was lower GLM scores.
It is embarrassing how much they suck up to american labs while those models only perform the first few days after release. With open weights you can be sure the quality stays the same unlike the closed source crap. How about a retest after some time?
The natural language choice in multilingual index seems pretty strange and biased to me, they tried to encompass a broad swath of Asian languages, but then there's like four European romance languages. They didn't include any Slavic one i.e Russian, or Nordic one like Swedish, or Finno-Ugric. And these romance ones are probably overrepresented in the training data anyways
I think it's pretty evident that Artificial Analysis loves to change the benchmark to place US models on the top, it's no the first time either.
The current placement of many models simply doesn't makes sense, GLM 5.3 is so much better than Astra at backend tasks, Astra does the most weird nonsensical bugs.
This is what happens when the benchmark team is funded by VC capital.
Even and especially for frontend UI, GLM 5.3 is a whole world better than Astra.
I can see (and believe) Astra beating GLM 5.3 on verifiable hard domains like maths, etc; but we’re at this stage where intelligence scores and even ‘frontier real world unsaturated benchmarks’ really doesn’t reflect real world use anymore.
Interesting that changes in the index always happen when „non us big tech“ models seem to be a little to good for comfort.
It makes sense for Muse Wonder though as that model does not trade blows with the best frontier models by a longshot.
agreed. It is not easy to benchmark these models. It just the date of index changes have some „spurious“ correlations
Open-weight models
Kimi K3.1 Pro: −9
DeepSeek V4.1 Flash 1017: −13
GLM-5.3: −13
MiMo-V2.6-Pro: −16
Proprietary models
GPT-6 Astra: +4
GPT-6.1 Sol: +3
GPT-6.1 Terra: +3
Claude Opus 5.5: 0
Claude Sonnet 5.5: −2
Claude Fable 5.1: −3
Gemini 4 Argon (high): −3
Muse Wonder Full: −6
Claude Haiku 5.5: −7
Muse Wonder: −9
Gemini 3.8 Flash (high): −10
Grok 4.7 (high): −11
Is there affiliation with OpenAI or coincidence?
Yeah, this smells. Ain't no way Terra went up and Deepseek so far low. I have had the later beat the former on multiple criteria by simply being less distracted.
There was a thread here about families of models responding in the same patterns to specific queries like "say a random word", and it takes less than two minutes to verify it is true. It'll take a bit of effort to not convince me this change wasn't reverse-benchmaxxed to screw everyone other than OpenAI. I'll be less distrustful after everyone is done IPO-ing.
My experience is not consistent with yours.
Well either AA is biased bribed riged whatever or chinese models are benchmaxxed or maybe its both
Neither is the primary cause, I suspect. If you look at the coding index alone, you see more plausible results from a software engineering perspective (in particular, Terra no longer outperforming Fable) [1].
That especially the small flash models that are primarily optimized for coding tasks would perform worse at non-coding tasks is not per se surprising to me. The introduction of Terminal-Bench Science 0.1 in particular hurts their aggregate scores [2].
The general question remains of how meaningful aggregating the results of very heterogeneous tasks into a single numeric score is. (Aside from general questions about individual benchmark validity.)
Note: the cyber index is also interesting. While refusals make it somewhat opaque, even if you optimistically count all refusals for Opus and Fable as successes, the results are ... fascinating [3].
I'll add that I really want to know how Opus 4.x and Codex 5.x perform on the new index. Once we have that information, this is going to give us a better understanding of how meaningful the scores are.
[1] https://artificialanalysis.ai/#coding-agents-tabs
[2] https://artificialanalysis.ai/evaluations/terminal-bench-sci...
[3] https://artificialanalysis.ai/#cyber-index-tabs
Tinfoil hat time: but the way they waited a couple weeks to evaluate Mimo 2.6 Flash, allowing GPT-6 Luna to sit on the Pareto frontier in the meantime, really did make me think they might have some bias going on... (Luna was displaced when they finally evaluated Mimo 2.6 Flash, which had been released the day before Luna)
edit to add: there totally is benchmaxxing going on, but I'd be really surprised if it wasn't across the board. After all, how can any lab tell if their model has improved over a previous iteration, if not by comparing benchmark performance?
I’ve very mixed feelings on Mimo. The new models do surprisingly well on some tasks, but they don’t allow you to set the thinking level and the model often times spends next to zero time thinking about a problem. If you coax it just a little bit, you can get it to solve it, but this isn’t an issue with other models. Also it’s incredibly easy to get it to get caught in a thinking loop.
> ...or chinese models are benchmaxxed...
It is hard to tell what people are trying to say when they use undefined words with no particular meaning. This doesn't seem to change the relative position of the Chinese models much. The Chinese are a month or few behind and the state of intelligence is so volatile that the top 10 will be completely different in 6 months. I'd expect it to still be a mix of US and Chinese models with the US models leading.
Benchmaxing is defined as targeted post-training attempts to optimise for higher benchmarks. This doesn’t change the aggregate intelligence or capability of the model, so it can make a model appear more capable than it is.
Glancing over the methodology[1] made me curious about three things:
1. Is there a benchmark measuring how much better an LLM performs when you tell it "you are an X"?
2. Does telling an AI agent "you are an AI agent" actually help? Is it the LLM equivalent of reminding a senior developer they're a senior developer, making them feel guilty about being lazy and finally getting them to do their job?
3. What happens when the persona conflicts with the task? If you tell it "you are a PM" and then ask it to write code, does it struggle? Produce worse code? Spawn a sub-agent and tell it they're a developer?
[1] https://artificialanalysis.ai/methodology/intelligence-bench...
Why did they call it “humanities last exam”?
Will an LLM topping this bench be the end of humanity? The doomer rhetoric has no place in what should be an objective measure.
"The FrontierCode verifier grades each attempt with the task's tests and rubric. We use GPT-6 Luna (medium) as the LLM judge"
Why are they using a closed source LLM as a judge? Closed source providers can see your traffic and adjust as they see fit.
I find it fascinating that some of the main areas people are interested in are using benchmarks with quite few questions: ~200 for coding and ~150 for science.
As long as that's the case, it's no surprise that there is often a perceived difference of quality between benchmarks and real-world use.
I know that running the full composite benchmark set is probably already quite costly, and also that the benchmarks have to be rotated regularly, but it feels like a significant untackled problem.
The drop for K3.1 Pro & MiMo V2.6 Pro is quite surprising. Are they benchmaxxed?
Or, are we witnessing reverse-benchmaxxing? Meaning that first some entities let players optimise for a given benchmark, then they update the measurement methodology so as to retain some players on the top ranks, while dropping others. As former Italian prime minister, and highly controversial public figure, Giulio Andreotti once remarked "Thinking the worst is a sin, but you’re often right."
Transparently so. MiMo's training dashboard showed the DeepSWE hillclimb in real time.
I keep seeing that MiMo is in the frontier in terms of performance and cost but I see very few talking about it and the only ones that do talk about it are doing it in context of AA
It's also dog slow at AA index tasks compared to other models in its class. Makes me think they optimised for getting on the intelligence/cost frontier at the cost of real world usability.
I’d bet on this
I used MiMo for a couple of days waiting for my Codex sub to reset. I really liked it, its communication style was a breath of fresh air compared to how stale gpt-sol is.
It's wild seeing opus 5.5 up there so high, while i'm still using Fable 5.1 because I couldn't stand how opus writes.
Opus 5 writes like its off it's meds, opus 5.5 is a lot less weird that way.
chinese labs are probably benchmaxxing but i'd like to see when the index changes relative to the position of the best chinese models (last suspiciously-timed update was when glm-5.3-flash was looking really strong)
This seems to be a hit and a miss, there's no visual indication that opus 5.5 is better even though in 9 tests out of 10 it will outperform astra especially when it comes to niche domains and will absolutely destroy them in anything requiring raw intelligence rather than a naive search, naive search is what makes openai models extremely capable in scientific fields though.
The score hit on smaller models is a welcome change especially when compared to models that nearly 10 times larger.
Grok being punished quite a lot is also a very welcome change as the model is frankly not that good.
I would love to see few more points knocked off haiku by introducing steerability/creativity benchmarks.
Honestly it does somewhat match what I’ve felt playing a lot with all OpenAI / Claude / deepseek models
One thing I’m still surprised / unsure on is haiku 5.5 max besting latest deepseek - but also I think the last haiku was so incredibly bad I still hold some bias whilst using the new one.
Interesting to see the open weight models drop so drastically from the changes here.
They perform pretty badly on the new benchmarks introduced.
[dead]
Where is GLM 5.3 Flash?
(EDIT: A child comment has refuted the AA score comments, AA runs using their own harness.)
One of the new benchmarks is FrontierCode 2.0 by Cognition. AA seems to have accepted Cognition's benchmark scores without questioning, but they should have looked more closely.
On this benchmark, Anthropic's models are run using Claude Code; OpenAI's using Codex, and Gemini is run using... Cognition's own Devin (not Antigravity). Thus its score is much lower than it should be, knocking ~3 points off its AA score. All open-source models are also inexplicably run using Devin, damaging their scores as well.
This may even be intentional, if Cognition has beef with Google, since Antigravity was built using Windsurf's technology. But aside from that, it is to their advantage to make Anthropic/OAI look like leaders in the AI market. They are a company known to lie and exaggerate again and again, so I have trouble giving them benefit of the doubt.
I don't think this is relevant on AA's FrontierCode score. AA runs FrontierCode 2.0 benchmark using their own Stirrup agent harness [1].
[1] https://artificialanalysis.ai/evaluations/frontiercode-2-0
My mistake, thank you for catching.
It's interesting that the Stirrup scores are somewhat close to the specialized harness scores. Either Stirrup is quite excellent or is it that harness doesn't matter so much after a point?
What's the advantage for cognition?
I would think it's the opposite as a more competitive frontier for llms makes the labs shed some margins and enable a broader ecosystem.
Yes, that is true, you might be right. But I think such, because the moat in their wrapper layer feels very thin and does not make sense in a world full of 20+ wrappers around similar models. It feels better to "only" have to deal with Ant/OAI (and other pure wrapper cos). Competing vs. Google, Meta, and every Chinese lab on earth, every one using their own bespoke wrappers, is much scarier.
Additionally: competitive frontier = more commodified = high valuations under more scrutiny.
Comparing the scoring rubric for AAv5.0 vs AAv4.3:
AAv5.0: 25% Knowledge work: AA-Briefcase v1.1 - 15% AutomationBench-AA - 5% GDP.pdf - 5% (-5%) 25% Coding: FrontierCode 2.0 - 15% (new) Terminal-Bench4.0 - 10% 25% Science: Terminal-Bench Science .1- 15% (new) CritPt v2 (hard subset) - 10% (new version) 25% General: AA-Omniscience v1.1 - 15% HLE-Diamond - 10%
AAv4.3: 30% Agents: AA-Briefcase v1.1 - 15% GDPval-AA v2 - 10% AutomationBench-AA - 5% 20% Coding: Terminal-Bench4.0 - 10% SciCode - 10% 30% General: AA-Omniscience (Accuracy) - 10% AA-Omniscience (Non-hallucination) - 5% GDP.pdf - 10% AA-LCR v1.1 - 5% 20% Scientific Reasoning: HLE - 10% CritPt 10%
tldr: gained Terminal-Bench Science and FrontierCode, dropped SciCode, GDPval-AA and AA-LCR. Lowered weight of GDP.pdf.
[dead]
[flagged]