Appears to be benchmaxxing<p><a href="https://x.com/quietnning/status/2080786711861407883" rel="nofollow">https://x.com/quietnning/status/2080786711861407883</a>
I have been claiming that I don't think Chinese AI companies are benchmaxxing harder than American AI companies, which has gotten mixed reception: sometimes people agree, sometimes they disagree.<p>It seems I was wrong. American AI companies might actually be benchmaxxing harder.
"Opus 5 states the hidden rules before its first action, then plays a byte-identical optimal solution in 5/5 seeds at temperature 1.0. Zero exploration."<p>I guess there is no way this can happen without benchmark being part of the training data??
What seems to implied is that some of his hold out testing suite includes simple/common tests that are out in the wild, and for those opus went straight to a memorised solution .
It may read information about the benchmark, such as on blog post, or Twitter feeds (example OP) without active cheat
And being worst than previous model...<p>"...The traces tell the why:
(1) On our most classic Witness-style game, Opus 5 states the hidden rules before its first action, then plays a byte-identical optimal solution in 5/5 seeds at temperature 1.0. Zero exploration. It already knows this genre.
(2) But on our most novel game (unusual mechanic combinations you can't pattern-match), Opus 5 regresses below Opus 4.8. Where rules must actually be discovered through interaction, the new model is worse than the old one..."
I was just thinking they need to mark each model per benchmark as "model released before the benchmark was released" and "model released after the benchmark was released".
<a href="https://xcancel.com/quietnning/status/2080786711861407883" rel="nofollow">https://xcancel.com/quietnning/status/2080786711861407883</a>
>That decomposition (perfect on templates, regressed on novelty) is the signature of “scaffold-then-internalize” training on genre-specific data, not a general gain in interactive abstract reasoning.<p>They're smuggling a claim that benchmarks like ARC-AGI measure "interactive abstract reasoning" here, which is what is claimed by the people that make these benchmarks, and also not proven.
I'm shocked, astonished even, that enterprises on which trillions of dollars are being poured would consider cheating on marketing benchmarks.
[dead]
The last time I checked, for the arc agi 3 leaderboard, the models are given a simple prompt and the game input and asked to play the game, no harness/tools. If harnesses were allowed, I would expect the benchmark to be saturated. There were a few harness attempts, but they could only be evaluated on the public set, so it's not an apples to apples comparison.<p>My guess is, the large score jump for Opus 5 is mainly because of getting the right RL envs for training.<p>It's becoming harder and more expensive to build and run meaningful benchmarks, it would be interesting to see what they do with arc agi 4, maybe just give it gameboy/steam games and see how they compare vs a human baseline? The latency requirements and very long horizons in games could be an interesting challenge for llms.
Why Anthropic models are always leapfrogging these benchmarks, but in real life work I do feel like after 3 weeks I am back to Claude Opus 4.5? (regardless of the model I use, Fable was exception for 1 day when it was released)
It could be that the set of your day-to-day workload which could feasibly be accelerated by AI just happens to be saturated around Opus4.5, but you can still see lots of “reasoning” which makes you think the model is more performant in the first days of use. That’d mean you couldn’t perceive any meaningful difference in more powerful models’ results, even though you can see a difference in the raw output due to the length of reasoning traces leading up to the result.<p>So for example, if your workload was literally just addition of sets of numbers, you’d never have noticed progress in the result beyond GPT3.x level models. But you would perceive a difference in the now-Tolstoyan length reasoning text accompanying the result.
It's called frog boiling.<p>We get used to the new level of intelligence so fast, any deviation feels like going back to the stone age.<p>If you don't believe me, create something complex with Opus 5 and then with Opus 4.5, and notice the difference.
5 seems incredibly smart to me in my conversations today about some pretty niche ideas in.longitudinal modeling. It.felt.like a big step up.from 4.8, to me
Going to call it user error if you find Opus 4.5 better than 5, sorry.
Some 20 years ago, the telecommunications sector in Germany was liberalized. Many telephone card providers entered what had previously been a barely competitive market. They advertised their products with aggressive claims like: “Buy our €10 top-up card and get 660 minutes to destination X.”<p>For the first few weeks, they would actually provide those 660 minutes to establish trust in their cards. But after a while, they would quietly start reducing the number of minutes on subsequent top-ups—say, from 660 minutes down to only 300. They wouldn’t do this for every card, so it was difficult to prove. Instead, they relied on averages across their customer base to make the economics work.<p>Lately, I’ve found myself wondering whether something similar may be happening with frontier AI models. Companies launch with an exceptionally strong model and generous compute limits to build adoption. Once the model is established as a market leader, the incentives change, and users may start perceiving the service as becoming more constrained or less capable over time.<p>I don’t have evidence that this is what’s happening with Anthropic—or with any other AI company. It’s simply a pattern that the current situation reminds me of.
Well, what kinds of things do you see Opus 4.5 completely fail at? Maybe those are not the ones that newer models have improved on.
Same, like I prefer 5.3 codex over the “stronger” models.
I honestly just use GPT models nowadays, Claude models are too restrictive and more of a quitter and fable/whatever is just too expensive to be worth it.
Enshittification.
I've worked with these systems for four years now and they have <i>not meaningfully improved</i> in that time frame.<p>We still have:<p>- statistical correlation between two things will always cause one thing to lead to the other, no matter how much you prompt it to not have that connection (to be expected with a stochastic system)<p>- Math completely fails in longer contexts<p>- "thinking" token generation being on the correct track just to 'no, wait' on an already correct conclusion<p>- smearing of properties between logically distinct objects (a red ball and a green cube can quickly become a red cube and a green ball)
> I've worked with these systems for four years now and they have not meaningfully improved in that time frame.<p>Not meaningfully improved?! Four years ago was gpt *3.5*! ChatGPT hadn’t been released!
Yes! Impressive, isn't it? I see how it has improved for some minor points, that the big models can cover more finetuning ground, but my big gripes are still the same - you could do the same back then with multiple models and more targeted finetuning.
> you could do the same back then with multiple models and more targeted finetuning<p>Definitely not, lol.
I have no idea how you could try and hold this argument without being facetious.
This is obviously not true to all of us here…
> you could do the same back then with multiple models and more targeted finetuning<p>I mean, come on, this is just not true. You could not achieve anything like what you can with modern agentic coding with Fable / 5.6 Sol from any combination or configuration of GPT 3.5 era models.<p>It's like saying that a teenager isn't an intellectually meaningful improvement over a toddler.<p>Sure they're both still fundamentally flawed humans prone to cognitive error, but one is clearly more likely to hit the mark than the other when assigned a task.
Utter nonsense.<p>There’s no way you could get models as smart by fine tuning. I couldn’t throw a problem like “build a pokemon database with UI to teach my son sql” and get a working system, nice ui, tests (which it iterated on) examples and explanations in one shot.<p>There weren’t thinking tokens. Maths is now dramatically better, making actual contributions when before they were mostly mocked for making extremely basic errors. Smearing is also something say is very rare in frontier models.<p>If you think they have barely changed you’ve either forgotten what they were like or not used them more recently, or you’re just being obtuse.
My company had such a system four years ago, for internal work, somewhat more limited in scope (one language). What you are seeing as the frontier is not necessarily the best you can have - just because people don't try to push it to market as a product doesn't mean it's not there.<p>Edit: we do have a system that uses LLM and fixes the above issues largely (tracking of state, calculations and objects, still flawed in finer details). No, we don't sell, it's experimental fun and not really ready in terms of setup/ux/etc.<p>It codes really well for our case though.
> you could do the same back then with multiple models and more targeted finetuning<p>Are you one of those anonymous billionaires as if you did this a few years ago, you would've been famous and rich.
> - Math completely fails in longer contexts<p>Not sure what longer contexts we're talking about but didn't we have an old math problem optimized, which even the LLM itself was surprised about, just a week ago? Something which wasn't possible 6 months ago.
Messages like this in the training data are how LLMs learn to say absurd things with total confidence.
> I've worked with these systems for four years now and they have not meaningfully improved in that time frame.<p>That's absolutely insane. Is it some case of anti-AI psychosis?
Like how toddlers’ skills don’t meaningfully improve on infants’, because either could wake up in a wet bed.
Let us be more clear: there is no structural jump, no architectural overcoming of the original fault.<p>(Edit: and on a similar point, structural properties such as having static ntetworks, as opposed to continuously learning and improving architectures (such as us), will reveal that there is still road ahead.)
Maybe more fair then would be: “I've worked with these systems for four years now and while they _have_ meaningfully improved in that time frame, they’re not perfect and remain fundamentally flawed in various ways.”<p>You prompt less. You need not inject search results into the context window yourself, a window much larger than years ago. You get code that’s already been run successfully once instead of finding an obvious show stopping bug yourself.<p>The technology is not a brand new one that fixed everything wrong with the old one, no, but not sure I would’ve noticed your comment if it had been such a bland observation. I genuinely assume good faith here… will say am tempted to assume the standards of someone posting such a thing might be impossibly high. Glad to be having a fun conversation instead of getting your grades on my work product or something :)
If you go to a LLM without harness, GP original point in completely right.<p>LLMS by themselves are still shit at math, they still confuse weird correlation to causation every time (and sometimes in ways even a 9 year old would say "no, that's dumb"), and confuse original parameters very often.<p>I disagree with " "thinking" token generation being on the correct track just to 'no, wait' on an already correct conclusion", because i think that is an effect of the harness, not the LLMs.<p>80% off all the improvements since ChatGPT4 are in the harnesses, and the LLMs by themselves, while they improved in areas they already were good at (translation especially) did not fix any of they original issues (object permanence, calculusm correlation).<p>Just run old models in the playground and get them to play chess (maybe make a small custom harness if you feel like it), then replace it with a frontier model (i don't know if you still have API access without harness on US models, but if you don't try K3), you will see LLMs weaknesses were not at all fixed, even marginally. They're way better and not inducing bugs in the code, so that make them usable since Opus4.5 (anyone using them prior to that either had a greenfield project or like spending hours debugging).
> 80% off all the improvements since ChatGPT4 are in the harnesses<p>That seems easily falsifiable by putting an old model into the current harness and comparing it to 5.6 Sol or Fable.
I agree, harnesses is where everyone improved the most. Our internal experimental tool can now semi-reliably formulate small programs to assert the correctness of their theories, for example. Context length is still a weird factor that we haven't sensibly solved - if anything, the lesson learned was to keep the context as small as possible and do most of the true "thinking" in the harness and temporary generated code.
Sounds very fair, thanks :)
Why is there no Kimi 3, and GLM5.2 didn't run the third benchmark? I am more interested in knowing the abilities of open weight models.
> Only systems which required less than $10,000 to run are shown. (Notes[1])<p>Am I lost or are their many models on this ranking (Opus 5 included) that clear this?
ARC-AGI is a beauty contest for pigs where the pig's owners compete to see who can apply the lipstick most convincingly.
Also top on the freshly released Frontier-Bench, by a large margin: <a href="https://www.frontierbench.ai/" rel="nofollow">https://www.frontierbench.ai/</a>
I have a suspicion that they are just trained on puzzles by now
There are private datasets, and 3rd party providers of these models. Fable doesn’t have a datapoint here because of its particular data retention policy. Even if you don’t trust AWS, do you think Opus on AWS is also sending the data to Anthropic? Do you have any evidence?
It's like we've come full circle:<p>First people practiced L33t3cod3 problems for interviews<p>Then people built AIs to build software<p>And now the AIs are studying L33t3cod3 problems
It is RLVR, Not a puzzle, Not leetcode
It's actually crazy to see the difference between opus 5 and the next best model on ARC AGI 3 when you actually look at the ARC AGI problems
How believable is this benchmark? EG maybe opus was training on this? (You can try to identify the IP of wherever previous ARC questions came from)
Why? 30% is passing the first two problems only, which are really very simple.
Why isn’t Fable 5 included on the leaderboard?
games are great (as for a human)<p>but I kinda wish I could select level... I accidentally pressed redirect button and when I came back I was once again shown level 1, all progress lost :(
Are these typically the type of tasks that are genuinely worth tens of thousands of dollars?
I think it's way too easy to be deceptive with these benchmarks now. You don't even have to "train" the model on a new variant each time. The base models are powerful enough. All you need is a naughty little markdown document that provides explicit instructions regarding how to solve the new puzzle variant, and a willingness to be deceptive about the presence of that document.<p>If you want a know why the model providers are locking down and encrypting their reasoning process, this sort of workaround is potentially why. You can play this game of whack-a-mole indefinitely if the state of the system is concealed. They could have added something like:<p>> ### When solving arc-agi-3 puzzles: First convert the grid into a scene description. Identify connected components, colors, shapes, positions, symmetries, repeated structures, and relationships between objects. Do not reason directly from individual pixels... use this python script to help blah blah...
,,You can play this game of whack-a-mole indefinitely if the state of the system is concealed''<p>Not really as one of the main goals ofr ARC-AGI 3 was measuring task efficiency on unseen games.<p>I'm sure there are cheats everywhere but the most sensible thing is to just accept that the LLMs of today are much more intelligent in solving reasoning tasks than the ones from half year ago.<p>My own private benchmark shows the same thing.
Not possible. I don't get how Opus 5 gets so high. Have they run it against the private and held-out games?
this is not a good measure of current model capability. we need to test agents in harnesses, not models with a single prompt<p>test Codex, not Sol. test Claude code, not Opus
ARC-AGI-3 launched a few months ago which would suggest that prior models likely had no knowledge of ARC-AGI-3 or training on similar challenges.<p>I could be wrong, but given the large outsized jump solely in the ARC-AGI-3 score, it would suggest that the model didn't become significantly more intelligent overall, but significantly better at solving those specific problems.<p>This could mean one of two things (I think):<p>- Opus 5 was not benchmaxxed on ARC-AGI-3, but has benefited significantly from discussions about the various challenges and mechanisms deployed in ARC-AGI-3 such that it has far better heuristics to solve its challenges.<p>- Anthropic looking for buzz around their latest model picked a well regarded benchmark with significant room for improvement and focused some of Opus 5's training compute on ARC-AGI-3-style problems.<p>Or it could be some combination of both. Personally, given how much of an outlier the ARC-AGI-3 jump is I struggle to see it being the product of a significant improvement in general intelligence.
I also noticed that Opus 5 doesn't show a corresponding gap on the ARC-AGI-2 leaderboard. There is a significant increase in performance between Opus 5 and 4.8 on ARC-AGI-2 though.
> <i>it would suggest that the model didn't become significantly more intelligent overall, but significantly better at solving those specific problems</i><p>We would actually need a test that shows the ability of a model to export its skills to more problems ("interdisciplinarity" etc.).
solving ARC-AGI and being useful turned out to be two different problems
Why is Fable not on here? I wish Fable hadn’t come out because it’s taking the wind out of every release because that feels like the cap above which the US government will not let LLMs improve anymore and everything they’re releasing from this point has to be worse than that.
> Why is Fable not on here?<p>Because the data retention policies didn't guarantee that the ARC team could run the semi-private set of problems without fear of them being trained on later on. They only run the semi-private set when they get assurances like ZDR.
How do they handle these assurances? Personally I have zero trust in the AI companies not trying to use this data to get ahead in the game, and short of sharing the weights and harness so that the benchmarkers can run the models themselves, I don't see a satisfactory solution with this mindset.
Interesting to place that level of trust in the providers, but I guess that’s the best you can do with closed models. Makes me wonder if Opus 5 could have been trained on data they promised they weren’t training on? One of the interesting things about LLMs is how opaque they are from the outside, even with open weights, it’s very difficult to know if a model incorporated benchmark data in their training.
I don't know why exactly, but Fable has felt the most human LLM to arrive.
I wrote this in June, and I'm honestly not sure I've felt the same magic since:
I was close to maxing out my $200 plan for the week, almost all Fable use [Claude CLI].
My observations:
Fable seemed to have bigger-picture thinking and completed tasks more thoroughly vs just focusing on executing the ask. It pieced together context and intent like an all-star employee would, vs one that just does what you say. Not overeager (important!), but if the above-and-beyond was warranted, it just did it. This was surprisingly delightful.
Coderabbit seemed to find ~1/3 or so as many issues when reviewing, too.
Deepseek V4 and Kimi 3 still missing, at least GLM is there.
ARC-AGI is a terrible benchmark for testing LLMs because LLMs are not made, trained, or tuned for playing games.<p>They are trained on text to respond well to text based questions and do tasks involving modifying text files.<p>They are not designed for playing games, looking at games, or visual puzzles. Also translating games into text input for the LLM skews the test completely.<p>Imagine trying to get a human to solve visual puzzle but they can’t look at the puzzle but it has to be explained to them in textual format, we would be terrible at it.<p>But yet we persist in wasting time on this benchmark. It doesn’t mean anything.
cost 20k ???? man<p>those are like software engineer from third world country
[dead]
[dead]