Qwen4-Max pelicans at four reasoning efforts
(none, low, high, xhigh): <a href="https://tools.simonwillison.net/markdown-svg-renderer?url=https%3A%2F%2Fgist.github.com%2Fsimonw%2Fff451620c173ea21e21b97b8cd46805b" rel="nofollow">https://tools.simonwillison.net/markdown-svg-renderer?url=ht...</a><p>For comparison, here's one I got from Qwen3.8 Max (at default reasoning effort): <a href="https://simonwillison.net/2026/Jul/20/afraid-of-chinese-models/" rel="nofollow">https://simonwillison.net/2026/Jul/20/afraid-of-chinese-mode...</a><p>This inspired me to re-run pelicans through 3.8 Max - but OpenRouter has qwen3.8-max-prime and qwen3.8-max-0902 - prime is twice the price and faster - so I went with the 0902 one.<p>Turns out Qwen4-Max has a none setting, but 3.8 Max didn't allow disabling reasoning - you had to pick from these five levels instead: minimal, low, medium, high, xhigh.<p>Here are those 3.8 Max 0902 pelicans: <a href="https://tools.simonwillison.net/markdown-svg-renderer?url=https%3A%2F%2Fgist.github.com%2Fsimonw%2F88071abe6ecd0f8e13ab77ade259be9e" rel="nofollow">https://tools.simonwillison.net/markdown-svg-renderer?url=ht...</a> - they're mostly not as good as 4, but I prefer the 3.8 xhigh one.
I don't fully understand why you keep doing these as-is. You original motivation was that this was not in the training set and indeed it was a great benchmark for a while, but it's clearly been incorporated now in the training data and possibly internal benchmarks.
Because he wants the engagement they generate. I don't understand why they still get upvoted on HN, but here we are. If even I can recognize that they're pretty much useless as benchmarks at this point, he must know it too.
He’s our pelican man :) I will always upvote pelican posts from the HN resident pelican guy
Likewise, i always upvote these posts even if i know it is biaised, it is always fun to see their evolutions, there is nothing wrong with them. Also even if we know there are trained on them specifically, some models are nonetheless better than others in engineering the output, thus it is always interesting.
You see it was funny the first and second time two years ago or so, now it's just boring. Whenever I want to read some insightful comments about a new model I need to scroll down through a few pages of comments about pelicans as if they meant anything now.
You can click the '-'/'next' symbol on a parent comment to collapse it. You are welcome :)
He seems to have posted it as a top-level comment, so it should be possible to fold it all, right? If he hadn’t, multiple other people might be posting about pelicans. Consider it the pelican attracting sacrifice post.
Things like "I will always upvote X posts from the HN resident X" feels like what HN should actively work against.
I have a colleague who got to the last rounds of interviews at a FANG, but did not make it...He could not draw a Pelican...
It's shame that people try to farm engagement on HN. Is this transforming to reddit. I don't know at this point but I agree with parent comment. EDIT: One thing I know is that I now need to scroll past the pelican thread if I need to read something usefull
It's part of culture at this point.
Based on some of the trash that consistently makes it to the front page I am very much ok with Pelican Man posting comments linking to his benchmark, even if its a bit less meaningful than it used to be.
And he spends the time doing it and as part of it drops nuggets of information not seen anywhere else. Respectfully and highly welcome. You're most welcome to move on.
Look at the page, though. If the data is in their training set, then the training must not have been very good, cause the images have very salient defects, and random artifacts popping up.<p>As far as I've seen, among the popular models today, there are still significant gaps in the quality of bike-riding pelikan svgs, making the benchmark still useful. Maybe not as a differentiator of frontier models, but certainly still informative for lesser competitors.
Unless the website is pulling hundreds of thousands of views per pelican, I sort of doubt any firm is actually training or post-training to optimize for this benchmark. They'll more generally optimize for better SVG rendering and I imagine these pelicans arose out of the annoyance we all had many months ago with how utterly bumbling and forgetful models used to be when completing tasks like that.
He can do whatever he wants, I personally think they are quite informative, albeit I don’t trust them that much either. It’s like a vague heuristic to say “this looks like an interesting model to test out”.
It's more a meme/influencer thing now.
> but it's clearly been incorporated now in the training data<p>You are not the first one to think of this. And Simon has thought of it as well, and commented on it several times.
Still, by now it's just more tradition than anything else. It's just kind of neat to have a consistent benchmark across many LLMs, even if the benchmark is completely gamed, but that is the nature of benchmarks.
Well IMO everyone can make their own benchmarks pretty easily, and that's what we should be doing.<p>Harder to train off all them.<p>Aka, let him have his fun, others are making benchmarks too
If this is that case it should be easy to demonstrate by asking current models for an svg of a rhino surfing. My guess is you would see a similar trend, given that the latest models are demonstrating large leaps in all sorts of content creation including 3D animation, rather than this being a demonstration of some sort of pelican-maxxing.
I love these. They're fun.
Agree.
I absolutely agree. its also fun to see the fun-hating gremlins emerge every time simon posts these.
OK, but why do you like fun, though?
They might be in training set, but there's still things wrong in them. That's not a great look for these companies IMO.
It keeps getting upvoted, so clearly a lot of HN users still find it valuable
Because it’s fun! The moment pelicans stop is a sad day. Keep them coming!
How can they not keep doing it? It's become unstoppable.
I feel bad for the data scientist that has to find or create images of pelicans riding bicycles for the training set to pass the meme benchmark.<p>Im not sure how you get the pelican to sit still let alone pedal, but there is a reason ai salaries are so high.
I wouldn’t be too sure.⁽¹⁾<p><i>“there’s little evidence that AI labs are pelicanmaxxing. Or at least they’re not doing it in a plainly obvious manner. Pelicans aren’t drawn any better than other animals. Bicycles aren’t drawn any better than other vehicles. And no lab draws the combination better than its pelicans and bicycles already predict.”</i><p>⁽¹⁾ <a href="https://dylancastillo.co/posts/pelicanmaxxing.html" rel="nofollow">https://dylancastillo.co/posts/pelicanmaxxing.html</a>
I feel like there's emotional attachment and influencer branding energy with this by now. It's more like a developer relations or developer ambassador type of content.
> ...but it's clearly been incorporated now in the training data...<p>It clearly hasn't; if it was quality of the pelicans and bikes would be static over time. Instead the quality has been improving along with the models.
Because it gets upvoted which means there is demand for it here
He should keep doing Pelicans, but every time throw in a surprise animal and a surprise locomotion vehicle. Would help identify the ones training for the Pelican.<p>Would love to see an SVG of an Elephant on a skateboard...
Someone, somewhere, at some AI lab, has by now said something like:<p>"We need to make the pelicans look good, but not so good that it looks like we're using them as an internal benchmark."
It's a little light hearted fun?
It's a fun tradition. Remember fun? We used to have that on the fucking internet.
Simon has clearly stated previously that he compares with other "animal doing human thing" type prompts to compare and see if theres a chance they are cheating by training on pelican svgs. Thus far, nobody has been caught cheating.
Actually, a Google employee has publicly mentioned that they’re specifically working on improving SVG generation in Gemini. Simon Willison has also pointed this out when evaluating Gemini models. [^1]<p>[^1] <a href="https://twitter.com/sunjiao123sun_/status/202455551655137292" rel="nofollow">https://twitter.com/sunjiao123sun_/status/202455551655137292</a>
I more don't understand why these comments get upvoted to top comment all the time.<p>Like... we get it. Simon likes pelicans. If I want to find his boondoggle robo pelican drawings, I know where to find them by now.<p>No need to spam every single model release thread.
You can click [-] to collapse thread.
Agreed. It's really annoying to come into the thread and a link to a pelican I have no interest in ever seeing is the top comment. Does anyone think these have any actual value any more?
Man, some of you people really have no whimsy :D
On the flip side, I love the pelican posts, but I find the people who think they're cleverly pointing out for the very first time that the benchmark may have been gamed by now tedious in the extreme.
At this point Simon can we maybe think about making this a 3d model? Maybe GTLF/GLB or custom Json you could even keep closed source for some kind of barrier to it getting trained on?<p>I say this because, even just based on my own product related to cad design/3d modeling, the SOTA models have gotten very good at 3d and I don't think the 2d SVG generation tells us much anymore. Maybe I can whip this up myself on my own site if anyone is interested tbh.
If you use the same prompt via the qwenapp or qwen studio you get an interactive web page with an animated scene. You can control every element of the scene, the colors, speed, time of day l, visible accessories etc and then download the one you like best.<p><a href="https://chat.qwen.ai/s/deploy/t_5fb28dd7-a61f-4946-8e59-8dd1c43d015d" rel="nofollow">https://chat.qwen.ai/s/deploy/t_5fb28dd7-a61f-4946-8e59-8dd1...</a>
I feel like the xhigh result is one of the first times I've seen a model go for a non-red colored bicycle (teal). Still haven't seen the leftward facing bike though.
Teal is a popular second color for Qwen Flash Next 3.8 as well. Also, I am not sure people realize how much variance there is for the same prompt. Enjoy.<p><a href="https://stealthgoat.com/fnfl/pelicans3.html" rel="nofollow">https://stealthgoat.com/fnfl/pelicans3.html</a><p><a href="https://stealthgoat.com/fnfl/pelicans8.html" rel="nofollow">https://stealthgoat.com/fnfl/pelicans8.html</a><p>My favorite, and shows this is not all benchmaxxing, this model is just good at SVG: <a href="https://stealthgoat.com/fnfl/pelicans6.html" rel="nofollow">https://stealthgoat.com/fnfl/pelicans6.html</a><p>Oh, and, quantization matters: <a href="https://stealthgoat.com/fnfl/pelicans2.html" rel="nofollow">https://stealthgoat.com/fnfl/pelicans2.html</a>
Astra and the OpenAI models do this color as well
Feels like Pelicanbench is getting saturated now. Maybe time to switch to a different animal and vehicle combo?
You say that, but it still takes quite a bit of reasoning to get the pelican's legs on different sides of the bike and to get the bike chain included.
The 3d one still produces interesting variations - <a href="https://www.reddit.com/r/ClaudeCode/comments/1xs4ll7/fable_52_vs_opus_vs_astra_pelican_riding_a/" rel="nofollow">https://www.reddit.com/r/ClaudeCode/comments/1xs4ll7/fable_5...</a>
Or something else entirely, but you'll still need the pelican for comparison, I would think.<p><a href="https://aishodo.enpitsu.org/?lang=en" rel="nofollow">https://aishodo.enpitsu.org/?lang=en</a>
Pogo-stick jellyfish, let’s go
First thing I do when a new model drop is scroll down to find the pelican comment, and then the "the pelican benchmark is no longer useful" debate under it.
What if it's been a Stork this whole time?
Kind of yes. Except from an artistic perspective they're all still awful.
I kinda like the no-effort result more than low, but high and xhigh return to what one would expect for reasoning->quality.
I don't understand why this is always the most upvoted thing for a new model. It's like some influencer "data" at this point. We're basically in 2027 now, the pelicans had their run, it's over.<p>Not a dig at you Simon, you've done good stuff, more just in general. Why is this always upvoted, it's like reading The Sun or something.<p>And why always one image. Should it not be say a 100 images then averaged or something? A prompt with temperature is a slot machine in Vegas, wtf does that have to do with model quality.
It’s a fun little exercise where the output roughly correlated with the progress of models.<p>The “real” benchmarks are widely available, and frankly kinda sketchy anyway, so why not have a little fun?
It's time to move from pelicans to horses.
You should call this The Pelican Brief
Man, how hard is it to understand that you straddle a bike, and one leg should be behind the bike? Apparently pretty hard, since all but the xhigh fail this ¯\_(ツ)_/¯
It has the legs on the proper sides of the frame only on max reasoning, which by coincidence was also the case with opus.<p>So either perspective is properly hard or the moonshine is strong in this one.
[dead]
It scores a 43 on Artificial Analysis with an average cost of $3.27 [1]. For reference, GPT-6.1 Sol (medium) gets a 46 with an average of $0.36.<p>[1]: <a href="https://artificialanalysis.ai/models/qwen4-max" rel="nofollow">https://artificialanalysis.ai/models/qwen4-max</a>
We should stop using this site. They now include private benchmarks which cannot be replicated. And every time an openai model falls down they change the index until it floats to the top again.
Terminal Bench is great, public evals, very complete and highly complex. I was browsing through them today and it gave me a real appreciation for the benchmarkers.
The key issue with public benchmarks is that they allow the model providers to train against them and artificially inflate their score accordingly. Private benchmarks are a way of making it harder to game the system, though it makes it harder for people looking at the benchmark to assess it as well.<p>Neither option is perfect.<p>A third option would be updating the benchmark every 6-12 months, but that has its challenges as well as you need to make sure that the benchmark covers things the model hasn't seen or is harder for the models to game.
It’s also worth keeping in mind the other alternative, having internal benchmarks that more accurately reflect your workload, that you know all about, and the LLM provider doesn’t. There doesn’t seem to be a way to maintain such a thing specifically for the purposes of online chatter (both because it isn’t profitable, and because things like “your workload” aren’t possible to define in that case). So we should be happy with what we’ve got.<p>This is a permanent problem in benchmarking really. Transparency is required for trust but invites over-tuning.
Makes me wonder how do you privately benchmark a model hosted by the lab. Even with zdr, the bytes are flowing through their infra and possibly influencing god knows what for the next training round.
You use something like item response theory [1] to be able to change the test over time while maintaining validity. It's what standardized tests like the GRE use.<p>Interestingly, Artificial Analysis does have a score that uses IRT, it's their Progress Index [2].<p>[1]: <a href="https://en.wikipedia.org/wiki/Item_response_theory" rel="nofollow">https://en.wikipedia.org/wiki/Item_response_theory</a><p>[2]: <a href="https://artificialanalysis.ai/methodology/progress-index" rel="nofollow">https://artificialanalysis.ai/methodology/progress-index</a>
this is an worry I sometimes see. I want to note that these are billion/trillion dollar companies we are talking about that have mutual talent movement (= uncontrolled information flow). Violating ZDR might be contractual violation, or at the very least, a strong breach of trust.<p>(not to say there is no value in "zero-trust" environments, but often pure-benchmaxxing is a deadend; some benchmaxxing is always involved and much less risky options than "train on the ZDR test env" usually exist)
Would OpenAI also "see" the request if it went through something like AWS Bedrock? This isn’t their infra and afaik the whole idea of this kind of intermediary.
The private benchmarks can be seen by the labs when you run them. We only have their word that they wont look at them, which isn't worth much. We should assume that everyone benchmaxxes equally, apart from a few outliers, and try to climb the hill by producing relevant training data.
Yes, that's another challenge with constructing benchmarks.<p>You could run the benchmarks using downstream providers like AWS if they are available.<p>You could vary the parameters of the tests to both help with LLMs learning as is and to help obscure that this is a test. Especially if sending them through a thirdparty, so the providers don't get identifying markers like IP addresses.<p>But ultimately it's another arms race like with viruses and anti-virus software.
Eh, I have an eval suite dating back to before GPT-5 in parts and have yet to see certain labs do as well on that as the public eval data would suggest. Course, I am pretty small, this is anecdotal and n=1, though I have seen very small, niche evals being clearly trained on the second they became public on GitHub for what that’s worth.
I think the best solution is really to update the benchmark periodically, and publish the benchmarks after the results are published.<p>Otherwise you have to rely completely on trust that AA is really measuring intelligence.
Or to separate the benchmark data by type and enable the user to filter by private, public or both.
Livebench does it
A fourth option is realizing that benchmarks exist because it is impossible to measure the value we want to measure, so we need a stand-in. The stand-in will sometimes correlate with performance but often just act as a way to compare. When used to compare, it is most important that the benchmark be fair and trustworthy.
I have to defend private benchmarks on this front, big believer in them, have seen how they can surface major performance differences.<p>Certain labs have been shown to and will continue to train on public evals, whether unintentional or on purpose, any respectable ranking across multiple evals also needs private data to counteract this.
We really should. I've been using other benchmarks like <a href="https://livebench.ai/#/?cats=Coding" rel="nofollow">https://livebench.ai/#/?cats=Coding</a> and it is more aligned to what it feels to my experience as well
This has Terra 6.1 at 3rd place on coding.<p>I have been using Terra a lot and I'd say that most of the models (even small ones) are are roughly equivalent when it comes to working in a consistent, established codebase.<p>But in practice, Terra is not very good at staying on task compared to say Sol 6.1 or Sonnet 5.5. Terra 6.1 will often just pause, thinking the objective has been met or the turn has ended.
I don't understand your point about openai. I can't remember the last time openAI was at the top of the leaderboard? Google had a brief moment at the top recently but other than that I feel like it's always Anthropic.<p>I kind of agree that closed benchmarks are bad. With a closed benchmark you still send the questions to the model provider and they may well train on it. Might as well publish it.
Even worse, it lags on my device!<p>Do you know a better site?
If they made the benchmarks public then Qwen would train on those to give the appearance their models are close to the American frontier. They have never been close. It's well established Chinese companies cheat the benchmarks which is why until recently they gave the appearance of being close to frontier but in actual real use they are far behind.<p>You can easily come up with a difficult private benchmark and test them yourself to validate the intelligence gap. It is real.<p>China lost. There is no catching up. Ever. That ship sailed.
private benchmarks are good
The table on that page shows gpt 6.1 sol (max) at $1.27 per task and it's the only model cheaper than Qwen 4 max.<p>By comparison opus 5.5 (max) is $10.01.<p>If they are not testing Chinese models with less than max thinking, then we should only compare them to other models also at max thinking.
> If they are not testing Chinese models with less than max thinking, then we should only compare them to other models also at max thinking.<p>This is wrong.<p>To make a vacuum analogy, one brand's Low setting could be another's Medium setting.<p>To compare you either give them the same resources ( Watts or $ ) and see how much they can do, or you give them a task and then check how many resources they required to finish it.<p>The label on the knob is irrelevant, except that either model might have a setting higher than the other which is sometimes useful if the resources aren't a constraint or lower when they are.
> To make a vacuum analogy, one brand's Low setting could be another's Medium setting<p>Which is why you need to test a range of settings.<p>Suppose you were testing car mileage. For the big name popular cars, you do 10km drive at 20km/hr, 40km/hr, 60km/hr and then publish the results.<p>For all the less popular cars you only do one test at 60km/hr and then you publish all the results for all cars on one big "Pareto frontier" graph claiming to show which car will allow you to drive home most efficiently.<p>It happens that lots of the less popular cars are not very efficient at 60km/hr so naturally the graph makes them look bad. Would they also be bad if tested at 20km/hr or 40km/hr?<p><i>Nobody knows</i>. Because this is where the analogy collapses - unlike cars, hardly anyone is doing these kind of tests for LLMs.
Well, sure, it would be better to evaluate all of them throughout the range. But it does not follow that we should further restrict the information we do have in this specific way:<p>> If they are not testing Chinese models with less than max thinking, then we should only compare them to other models also at max thinking.<p>If we have data only for “Chinese max” (let’s say it costs $1) and “non-Chinese low-to-max” (spanning $0.10 to $3), comparing them in full at least tells us what we can expect from each model if we have $1 to spend + the maximum performance we can expect with unlimited budget. What’s the point of restricting to “max vs max”, which answers only the second question?
Usage? If most people predominantly default to Medium thinking level for usage then it would make sense to benchmark that first.<p>Another analogy for this are review sites overly focusing on reviewing all the SKUs of GPUs that cost $1k+ for an audience that, according to Steam's Hardware Survey, predominantly use midrange X060s-equivalent cards. Sure, it's fun to know whether or not you can enable PT on a 5090 but for most of us, covering 4070 TIs and below in-depth (meaning actually taking the card apart, seeing how good/bad the past jobs are, possibly buying multiples of the same SKUs to evaluate QC, etc) is much more useful because of the value prop.
> Usage? If most people predominantly default to Medium thinking level for usage then it would make sense to benchmark that first.<p>Sure, that’s not in question. The discussion here is not about what to measure (we agree that benchmarking only max is suboptimal) but how to compare models if you have an already fixed set of measurements, outside of your control, which includes only “max” for model A but a wider range from model B.<p>The person I was responding to was claiming (in their previous comment) that the sensible thing to do, then, was to artificially truncate the data for model B to “only max” as well.
sounds like context collapse
This is a silly take. To continue the analogy vacuum A might need the power at 50% to vacuum something, B 100%. You cannot then test both at 100% and say "both finished the task, but A required twice the resources so it's worse", since you didn't check what resources it <i>actually</i> needed.
The Artificial Analysis charts show max effort by default, which isn't how most people use them.<p>Scroll down to the cost per task pareto front chart which includes more difficult levels by default: <a href="https://artificialanalysis.ai/models/qwen4-max#intelligence-per-task-tabs" rel="nofollow">https://artificialanalysis.ai/models/qwen4-max#intelligence-...</a><p>Qwen4-Max is not in a good spot on that chart. Astra xhigh and Fable 5.2 medium had a similar cost per task with higher intelligence ratings.
I cannot see cost per task for any other thinking levels for Qwen 4 Max.<p>If it's actually there, what is the cost at medium thinking? It would show Qwen 4 Max (medium) if it's there.<p>Everywhere that I can see on the page it just shows ”Qwen 4 Max" while for 6.1 Sol, Opus 5.5 there's always a thinking level in brackets after: (low), (medium), (max) etc.
They don't always run the benchmarks at every reasoning level for every model; only a few prominent ones. They do it for the US frontier labs, plus a few of the Chinese labs, but not consistently. You can see that DeepSeek has a few reasoning levels tested, GLM 5.3 has a couple, as does Qwen 3.8 27B, but not other Qwen models, and a lot of the Chinese models they only run at max reasoning.<p>It's possible that they will run Qwen 4 Max at other reasoning levels in the future, or it's possible that they never will. Regardless, it's unlikely to be better at any reasoning level on a price/performance basis than the GPT 6.1 Sol/GPT 6 Astra pareto frontier. It's close enough to Claude Opus 5.5 line at lower reasoning levels that it could potentially cross that if it doesn't loose too much performance at lower reasoning levels. But it's far enough from the global pareto frontier that it's unlikely that it would cross that at any resoning level.
Right, so my original point stands: if they're testing this model at only max thinking level, then it's only valid to compare it to other models also at max thinking level.<p>You can see for example that the cost per task for 6.1 sol is about 4x higher at max compared to medium thinking level.
No, it's only valid to compare to models at the same cost level, regardless if that's their max level or not. Fable/Astra having higher effort modes than Qwen doesn't mean that you can't compare Fable medium to Qwen max.<p>Which effort levels are cost-equivalent is debatable and depends on whether you are paying API prices or on a subsidised plan or are able to run the models yourself on your own hardware. But I think AA uses public API pricing for everyone as a sensible baseline.
If you look at the pareto chart for e.g. 6.1 sol you'll see a curve of about 6 data points, some of which are on the Pareto frontier, some are not. For Qwen 4 max, many others, there's just a single data point so we can't just how close <i>any</i> thinking level would get to the Pareto frontier.<p>I'm not saying this model would land on the frontier - from the single data point it looks unlikely. However, that single data point might be making it look much further from the frontier than it should be.<p>EDIT: I will also take back my point that we should be comparing models at the same thinking level - max on one model won't correspond to max on another. What we need is a range for each model. What we have is a range for the most popular models, showing that the same model with different thinking levels can have vastly different results. Then for less popular models we have just one data point, which isn't enough to draw the kind of conclusions people are drawing in this thread.
> You can see for example that the cost per task for 6.1 sol is about 4x higher at max compared to medium thinking level.<p>It does not (universally) hold for long running workloads -- which is what people are increasingly doing.<p>Leave a Sonnet on medium, it will complete the task, spend more time and possibly significantly more token money, vs. Opus 5.5 on xhigh.
> The Artificial Analysis charts show max effort by default, which isn't how most people use them.<p>I always use Max effort in both Opus and Fable. It's actually faster overall because tasks get done faster with fewer "corrections" from my end.<p>On GPT models, I'd sometimes dabble in xhigh/high but that's because OpenAI usage quota are too limited.
How “safe” is Qwen4? Will it immediately shut down on security research like the US frontier models?
I assume those cost comparisons are on API token cost.<p>I'd love a comparison on for example monthly quota usage per task, between all ~$20/month subscriptions.<p>It's so difficult to compare how much "productive" usage you get out of each subscription, without signing up to each one at least.
> If they are not testing Chinese models with less than max thinking, then we should only compare them to other models also at max thinking.<p>Nah, there are interesting conclusions to be had in comparing various model/effort level combinations, esp. when it comes to cost and time per task.<p>I have my own private benchmark suite; it involves algorithmic problems, the need for good visual understanding, and the ability to run long workloads unattended (as the agents need to run a few iterations, self-evaluate the results, and hand over the assignments when they feel they have succeeded). Chinese models do poorly on such problems, and running them at anything less than max efforts is pointless; they can't even hand over the assignments graded against a mechanical set of tests they themselves can execute. Hand that same task over to, say, Sonnet medium and you have a weird outcome: excellent result, but way more time and tokens burned than, say, Opus 5.5 or Sol 6.1 running at high. I do those benchmarks just to be able to spot such patterns and then better choose the models for my real work - but i wouldn't have caught it if I wasnt running various model/effort combos.
> If they are not testing Chinese models with less than max thinking, then we should only compare them to other models also at max thinking.<p>They compare different reasoning levels.<p><a href="https://artificialanalysis.ai/models/deepseek-v4-1-flash-1017" rel="nofollow">https://artificialanalysis.ai/models/deepseek-v4-1-flash-101...</a>
Does anyone really believe the cost data for benchmarks? Pretty sure all models will behave differently at this point when they get benchmark shaped questions.<p>(And I’m sitting on data showing Sol 6.1 and Qwen 3.8 Max cost roughly the same per task, for real tasks.)
The cost per task metric is accurate for those benchmark tasks. If your tasks don’t look like benchmark tasks..you should use your own benchmarks :)
I mean not believing intelligence score is reasonable but cost is just reporting what they spent?
vals.ai puts it at just about Astra level fwiw.
Looks like a slightly weaker, cheaper Kimi.
I wish they'd also post the benchmarks according to the previous "version" of artificial analysis.
Artificial Analysis is a VC-backed firm and it's unclear what the incentives are there.<p>They have altered their benchmark multiple times in the last 4 weeks (v5 in <a href="https://artificialanalysis.ai/articles/artificial-analysis-intelligence-index-v5" rel="nofollow">https://artificialanalysis.ai/articles/artificial-analysis-i...</a>) and the last 5.0.1 change <a href="https://artificialanalysis.ai/methodology/intelligence-benchmarking" rel="nofollow">https://artificialanalysis.ai/methodology/intelligence-bench...</a> pushed up Anthropic's models a bit, you can see that in v5.0.0 at the first link, Opus 5.5 has a lower score than in it gets in v5.0.1, and v5.0.1 was released just before Fable dropped.<p>Some part of me believes that the benchmark mix was put together to favor Anthropic's models.
> Favor Anthropic<p>Rarely seems to make them look good whenever I drop in to the site. Far more expensive for little gain. Even within their own effort levels it looks bad, Fable xhigh is 50% more in cost than high for zero improvement across a range of tasks.<p>A tiny bump in intelligence score hardly seems worth it for cost, and honestly most people's tasks are not cutting edge work.<p>Current benches on there seem to make Sol look the best, at a certain point the subscription subsidy tap will turn off and people will be in for a rough time paying full price.
If Artificial Analysis is not to be trusted, then which benchmarks are?
I don't think there are any that can be trusted. There's too much riding on these benchmark results so the labs find ways to game them either by benchmaxing or by influencing the format of the benchmark.
Check others as well. <a href="https://www.vals.ai/" rel="nofollow">https://www.vals.ai/</a> has it just under Astra.
I think you mean that benchmaxxing has turned into a business model, not for the Chinese models (maybe them too, of course), but for the AI labs, and artificial-analysis is not remotely the worst offender.<p>Right now the big labs are buying benchmark tasks from "benchmark labs" (or whatever they call themselves) for absurd amounts of money, and making sure they perform well on them.<p>The incentives this generates are <i>terrible</i>.<p>The big AI labs will only buy benchmark tasks that the frontier models perform poorly on (because they have trouble finding any public ones). So the tasks are being mass-written now directly based on the models they're supposed to train. Obviously the math says you shouldn't do this. If there is a causal connection between the benchmark and the model, you lose the normal guarantee that improving benchmark scores will improve model.<p>In fact improving such benchmarks may make the model worse. To understand that let's just give a ridiculous example. I create tasks like "1 3 5 11 13 17, give next number". The next number is 301. These are the house numbers on some random street in Brussels, Belgium (I'm guessing the big jump is because they "merged" 2 streets, and just added 300 to the numbers of the second part).<p>If the model trains on this, performance will actually decrease because obviously this is not how you <i>should</i> reason.<p>BUT making tasks under this "impossible to reason towards" part is of course how you make tasks that might even be reasonable, and of course frontier models are going to suck on tasks like this. So they sell!<p>BUT they suck! Finding houses is actually going to work a lot better if you ignore this fact until it becomes unavoidable in one out of 1000+ streets.<p>And yet this is what benchmark improvement has moved to. If you look at the terminal-bench 4.0 tasks you will see what I mean. Doing better on those tasks ... does not actually make a better model BUT you have to actually go read those tasks to realize that. A lot of the new tasks largely demand you make weird jumps in your reasoning not based on understanding the underlying theory. They rely just being able to predict what goes wrong in certain cases when theory would suggest it should work, and mostly it even <i>does</i> work but for some reason not acceptably. They reward not doing the full work, realizing what goes wrong and correcting, without understanding why the normal reasoning won't work and correcting without knowing why. I find it very hard to believe this makes a model more intelligent. It rewards knowing the encyclopedia by hard, so to speak.<p>But it certainly makes it look better. Whereas the obvious, coding improvements, do not make models look better anymore.
AA must be doing something right. Ever since they released the new benchmark, you see open weight China model fans complaining about it under every release thread. Probably because it’s hard to benchmaxxed against.<p>But admitting their model is weaker is a no go. So conspiracy theory here we go
For what it's worth, in our own metrics and benchmarks the intelligence is a pretty good match, but the cost per task is wildly out of whack. We see the open weight models like GLM-5.3-flash or Deepseek 4.1 Flash 1017 use a fraction of their tokens/cost of the benchmarks, which changes the cost-benefit equation a lot.
or maybe it's the other way round, open weight models are under siege while closed weight models are defended
> Probably because it’s hard to benchmaxxed against.<p>It's mainly just a composite of existing benchmarks. It would be very easy to benchmax if, say, AA shared with a preferred lab which benchmarks it plans to upweight/downweight in the next version. Just speaking hypothetically, of course.
Yeah, it's in the same ballpark as Kimi 3.1. Both of which are obviously open weight, so that's at least a big differentiator from Sol. Let's hope against hope that the much smaller flash model might beat deepseek v4.1 flash and run for as cheap.
Eh, as long as they make qwen3.8-flash-next equivalent on less resources and faster, they're winning.<p>It's like arguing windows 11 just needs more bloatware.<p>These models don't have to all in one every task in the world.
Why do we trust this? Do they open-source their test code?<p>Also, a generic benchmark might not translate well into what everyone needs. I keep building my own test cases that test models for my own work specifics. Only that benchmark is useful to me. I don't need it to win a PhD math quiz.
It looks pretty bad on those benchmarks. Is it possible that this is due to the recent crackdown on distillation?
[dead]
Anthropic deserves to go bankrupt for trying to ban open-weight/open-source competition. I hope their IPO flops.
Anthropic and its CEO are preaching about “safety”. Meanwhile, they created a model considered “too risky” and gave US government access to it. This is a government whose track record includes detonating nuclear weapons over cities full of civilians, deliberately exposing prisoners to sexually transmitted infections, and dumping weedkiller across Vietnam on such a scale that it caused over 100,000 birth defects.<p>The US government is the most reckless actor on the planet and Dario just couldn’t wait to give it a tool to create new diseases, new atomic weapons, new spying tools, new biological weapons, and whatnot.
> The Army released S. marcescens in eight tests to determine the vulnerability to enemy attacks (U.S. Army 1977). In addition, the Army conducted field testing with B. globigii and Aspergillus. S. marcescens and Bacillus subtilis were also released in San Francisco and New York subway systems, respectively, to study their dispersion.<p><a href="https://www.ncbi.nlm.nih.gov/books/NBK233505/pdf/Bookshelf_NBK233505.pdf" rel="nofollow">https://www.ncbi.nlm.nih.gov/books/NBK233505/pdf/Bookshelf_N...</a><p>They also previously released bacteria from planes over their own citizens:<p><a href="https://en.wikipedia.org/wiki/Operation_Sea-Spray" rel="nofollow">https://en.wikipedia.org/wiki/Operation_Sea-Spray</a><p>See also: Similar "experiments" in the UK around the same time.
Just stating facts. I don't know why you're getting downvoted.
Biased, only shows bad examples without good examples the US government does.
The reply was in context of the post above theirs.
So those are okay to do, even to you, if watered down with enough "good" on top ? You sound like a partisan of the Leopard's Eating People Face Party. /s
"I stopped wearing my seatbelt after reading about people who lived because they were thrown from their car."<p>A pile of facts don't inherently build up to the truth. The sooner people learn this, the sooner we can leave behind so much bullshit and misdirection in society.
This is a government whose track record includes helping liberate Europe from the Nazis, rebuilding devastated countries through the Marshall Plan, putting human beings on the Moon, and funding scientific research that helped make the internet possible. Through PEPFAR, it has helped save millions of people from HIV/AIDS.<p>The US government has been one of the most consequential forces for human progress in history. Giving it access to powerful AI could help accelerate medical research, strengthen public services, protect democratic allies, and make scientific breakthroughs that private companies would never pursue alone. Dario is right to recognize the potential: helping democratic government use extraordinary technology responsibly is an ambition worth having.
Don’t let the fact that they never did that and have explicitly said they’re opposed to it get in the way of your anger.
They have been trying to do that. They never signed the open weights support letter and have been yapping about dangers and standards basically trying to shape the ecosystem on their view of how it should be; which would effectively ban open weight models.
<a href="https://www.implicator.ai/openai-anthropic-lobby-washington-open-weight-ai/" rel="nofollow">https://www.implicator.ai/openai-anthropic-lobby-washington-...</a><p><a href="https://thenewstack.io/microsoft-nvidia-meta-and-open-weights/" rel="nofollow">https://thenewstack.io/microsoft-nvidia-meta-and-open-weight...</a><p>Then there's all the "open models are dangerous" rhetoric while Anthropic and OpenAI's models are displaying said danger. Then there's the insane amount of money they're spending on lobbying. On what exactly? Who knows, but since it's their interest to kill open weight models and most companies tend to lobby to defend their own interests, it's not an unreasonable assumption that some effort is going into that.
When did they say they oppose it? I've only seen them actively scare mongering that open weight models are too dangerous?
I hope they are able to serve 4 faster. Qwen 3.8 is a great model but I’ve regretted including it in the battle system for opencode [1] we have been building because it scores so low.<p>And it’s a shame because the only reason it scores so low is that it’s slooooow. (Like 30 tokens per second vs opus near 80.) If you exclude voting, Qwen 3.8 does well.<p>(We allow early voting contrary to Arena.ai because I think it makes sense for agents, where it doesn’t for chat. This means our data takes speed into account.)<p>Also, it would be ideal if they served this with third parties. Part of the slowness with 3.8 is just not a lot of providers. Their open source version is much faster, but we couldn’t include it because it doesn’t handle images (another own goal by Alibaba).<p>[1]: <a href="https://agentduel.com/leaderboard" rel="nofollow">https://agentduel.com/leaderboard</a>
In case it translates to Qwen 4 27B, I've been running Qwen 3.8 Flash Next (which while MoE (and bigger) is supposed to be a preview of Qwen 4's architecture) and it runs pretty solidly regardless of context window size and completes random deepswe challenges I throw at it locally (at Q2-Q3 quantization! so it fits on my shitty setup).
This, I continue to be amazed using Strata to serve IQ3_XXS. It's such a capable model for something that runs on (relatively) little hardware.
That is our experience as well at FP8 for 3.8 FN.<p>For many of our use cases it's currently far into the "good enough" range because external factors (such as process design) are now blockers.
> but we couldn’t include it because it doesn’t handle images<p>Maybe I'm mixing up model variants, but Qwen 3.8 27B that I run locally totally does handle images. Surely the bigger one would as well?
The model I’m referring to that would be faster and similar size, but doesn’t handle images is Qwen 3.8 2.4T: <a href="https://openrouter.ai/qwen/qwen3.8-2.4t-a95b" rel="nofollow">https://openrouter.ai/qwen/qwen3.8-2.4t-a95b</a>. It’s weird that this is pretty much the same as Max but doesn’t support images. No idea why.
Qwen3.8 doesn't handle images?
Is the speed due to a difference in the way it "reasons"? I was trying out a local version (3.8 27b) and watching the thinking, it's absurd how often it challenges itself. Simple facts get questioned multiple times, and it has to keep reminding itself that it already established some things as facts earlier.<p>Instructing it to ask any and all questions as necessary sends it into an existential crisis, because the question tool (oh-my-pi harness) suggested limiting the number of questions, and it couldn't figure out how to do both.
Aside from making our engineering prompt terse, I place a token budget on the thinking traces. Those traces are mostly meaningless(in the reasoning realm, not the usefulness realm), and just tell it to do something productive. For coding, this just makes it read doc, tests, etc.<p>I believe that's all reasoning levels do, but most inference engines should have thinking budget.
[dead]
I think Alibaba is not aiming for the frontier on this class. Their previous Qwen3-Max was released in September, right around when Opus 5.5 was coming out, and it was nowhere close to it as well. However, one or two tiers below (Flash, 27B) they really shine and are SOTA or close to it.<p>DeepSeek has a similar problem, I feel like their full model is hardly even worth using, while the Flash tier is absolutely amazing.
It seems to be the case. I would qualify them more as flagship-killer than frontier, using the same terminology used for smartphone classification. As opposed to the big American shops which focuses solely on performance, they have operational costs and other factors in their constraints and requirements. As soon as the market reaches some maturity, this will prove to be very important.<p>Crisis sold far fewer copies than Call of Duty.
I am unsure. Why shouldn’t they? Why even work on a big model (especially training time) if “nobody will use it as competitors are better”? We’re still talking about nearly 2.5T params here. Deepseek’s pro model is still heavily smaller and like you said isn’t directly worth using, obviously it has the open source / self hosting advantage.<p>Imo there’s a need and they try to push here, it’s also giving them advantages for their smaller models but if the big models never perform they will have a bad time or maybe should even stop.
If you're aiming to make the best possible small model, you still need to train a larger model and then do logit-level distillation from it. This is where most of the gains in small models have come from.
My use of qwen3.8-flash-next suggests they're trying to make a low cost model that knows how to program and they think the "total addressable market" is serving more people at lowest inference costs possible.<p>Think of it like a hind brain+fore brain discussion. Instead of focusing on making some ultra model that can do everything but costs a ton, you have your smaller models gather the bits and feed it better data<p>Agentic coding is the target then. If they want their work force to not care about token gen, that's what you do.<p>As an analog, they're making a drone swarm while US is making 5th Gen fighters. Completely different battle fields.
Yeah, I feel like models like Qwen3.8 27B are making clear trade-offs, like having very little world knowledge baked in. I believe those are intentional, so they can focus more on other areas. I guess that penalizes them on these benchmark aggregators.
The 'frontier' incorporates two dimensions though, usually score x cost, so it's not just the best score.<p>Meaning models below the frontier either cost more for same score or score worse for same cost.
I think the argument is like MacOS vs windows vs linux. AI models from anthropic/openAI are windows and MacOS; they're plat forms for "skilless" to start on and of course you can gravitate towards the development end. They're bloated in that sense.<p>Open models follow the linux goals which is small tools that chain together. If I can get a small model to understand how to program for another model or do the programming bits for a large model, then the trade off is flexibility. You can imagine your giant AI that can do anything, like windows/Mac is to your benefit, but there's a clear flexibility line to how these models act.<p>Even if it takes longer for my local model to find something, that doesn't mean it's taking a long time because it's doing wrong things.<p>Honestly, I think a lot of people using SOTA don't understand what you can do with a proper engineering harness; we all live in a yellow submarine.
This is exciting. The continued focus on efficiency alongside quality is encouraging - it makes me hopeful to see what smaller models will become capable of.<p>Looking forward to the "Flash" and 27B versions.
Waiting for the Flash tier too. My experience running coding agents on cheap vs big models: the cheap one gets the same work done, roughly twice as slow. The big model buys you speed, not results.
I'm also hoping someone puts out a competitive 30-70B model soon. There's this massive dead zone between 24GB and <100GB of VRAM where it seems your best option is Qwen 3.8 27B.
The 3.8 Next-Flash release is comfortably the best model on a 96GB Mac Studio. I’ve tried various engines and Sushi (<a href="https://github.com/beamivalice/sushi" rel="nofollow">https://github.com/beamivalice/sushi</a>) is the most performant so far.<p>Hopefully the Qwen 4 Flash model will still fit within 96GB without requiring excessive quantisation as, verbose thinking aside, 3.8 is a very solid workhorse model.
There are many engines that alleviate this issue in particular. The current mood is running qwen flash next using strata, basalt, ect.
Same, I recently found Nemotron 3.5 74B-A9B really good at filling this gap though, could do (very slow) local inference on just 32GB DDR4, without the quality dropping to the floor
Quite a few saying that they tried earlier snapshots of this model and are finding it does not live up to expectations. Firstly do other agree and also why would the model get worse - could they be doing something weird to the model that is making less aligned at the end of training, potentially for political reasons?
I subscribed to the Lite plan, but I'm not finding where I can get/generate an API key, to be able to use it with something like Hermes Agent?
is the open weights lag widening? these open models don't seem to be improving as quickly as ant/oai
I think looking at the lag between frontier and open-weight is looking at an uphill battle and assuming it's equal footing.<p>Lots of research goes into open-weights, optimizing cache re-use, storage, inference costs, etc., and all these benefits are easily rolled into closed AI models. Meanwhile any advances made by OpenAI, Anthropic, etc. are kept hidden away to give their models an edge.<p>It'd already be a miracle if open-weights were gaining on-par with closed-weight models. That they've been closing the gap for most of the past year is wild!
No idea, but this is a rapidly changing field. Opus 4.5 was released a little less than a year ago and it feels shockingly limited compared to even current open flash models. (I literally just tested it the other day.)<p>So, with new models coming out on a weekly basis, I think trying to figure out a future trajectory is just going to be a wild guess. For all we know, Anthropic and OpenAI models may also be plateauing in the near future; or there may be a breakthrough and they may improve beyond our wildest expectations. I honestly don't know.<p>That said, I suspect available compute and model sizes that major Western vs. Chinese labs can handle will still keep the gap open at least in the short term.<p>On the other hand, I am also not sure how much the gap will matter past a certain point. I am specifically wondering at what point people will become interested in smaller models fine-tuned for specific application areas (e.g. cheminformatics) rather than massive generalist models.
I can't imagine them keeping up long term given the information asymmetry plus compute restrictions on top of it.<p>I feel open-weight was able to keep up the past year using efficiency improvements (hybrid attention, improved MTP drafters, improved optimizers) to offset having less compute. But OpenAI/Anthropic started cutting token costs just a few months after these, which to me suggests they copied these improvements. Indeed it would be hard to work in LLM training and not know about DSpark, Muon etc.<p>So what do you do if closed source uses your learnings, gives no knowledge back, and has 10x more compute and funding on top of it? The situation feels quite dim right now, esp. with the Nvidia thing.
open always wins
Right now open is heavily subsidized by China and I fail to see what would happen if that stopped.
What advantage do they have over closed? You may want open to win, but logically I don’t see how it would
I can run an open model on my own hardware, in my own home (which has solar panels on the roof, so electricity usage is mostly a non-issue). Or, if needed, I can rent compute from somewhere like you'd do for any other workload.<p>Once I have a model locally, I know it isn't going to change on me. Compare that to using an API from Anthropic or OpenAI (and yes, the only thing you can say is you're using an API, not one model), they can discontinue a model you're relying on (as every model could have peculiarities you've designed around) or modulate their speed/quality day-to-day to save on compute costs.<p>And trying to claim that buying your own hardware doesn't pay off compared to API or subscription costs doesn't hold much water either, at least not for me. For starters, I already had the GPUs and machines I started running models with, whether it was for gaming or for other reasons. And in the case of the new GPUs I bought recently (2x AMD Radeon Pro R9700; no you do not need CUDA and you haven't for a long time now), I can hold a reasonable expectation that they will hold at least some of their value years from now, compared to paying for API/subscription usage for years in which case you're left with nothing.<p>So, if you have a local model that is "good enough" (and they already exist for many use cases), it's just kind of the obvious choice to make, IMO. And in just one person doing it as a hobby while I'm still looking for a job 16+ months after a layoff. A small business and certainly a large enterprise has even less reason to open themselves up to all the downsides that come with using Anthropic or OpenAI.
Local models already crossed the threshold where they're quite good and useful for many things, even with cheap setups. I walked away really impressed with (non-coding) use of Byteshape's Qwen3.5-9B quant on my 8GB RTX 3070 for instance. The model was fast, helpful, and 100% private.<p>If we extrapolate over the next several years, prices for VRAM/RAM will start to come down as manufacturing capacity comes online, models will continue to become more token efficient, and intelligence will improve at every quant size. As this happens, more and more people will be running LLMs locally on hardware they already own, and there's going to be a point where local models can do everything you need to do unless you're literally trying to do breakthrough mathematics research.
I don't think "useful for many things" qualifies as "winning". It's about two orders of magnitude shy of winning.<p>> and there's going to be a point where local models can do everything you need to do unless you're literally trying to do breakthrough mathematics research<p>Maybe. But competitive markets pay exponentially more for incrementally superior services. Professional athletes are a great example of this. If you're building a product with competitors and those competitors have frontier models and you stick to local then you're probably going to lose both today and in the near-ish future.<p>If you're bitter lesson pilled then the alpha is entirely in having a better model, more data, and more compute. Your vision is that human ingenuity will provide more alpha. Oof. I wish. But my tastebuds are very very bitter.
> If you're building a product with competitors and those competitors have frontier models and you stick to local then you're probably going to lose both today and in the near-ish future.<p>i highly doubt this premise. the success of a product has very little to do with the amount of the intelligence used in producing it. you could throw millions of dollars using AI to develop a movie and it would still flop in comparison to an indie-made film.
Also - the “open” here doesn’t have the advantage of open source, where it can be improved by the community. At best, the community can tweak it a little. For actual improvements you need to wait for the next model which is trained by companies for billion of dollars, and could stop releasing them at any moment.
It can, through better quantizations and reinforcement learning.<p>Just this week, qwen 3.8-27b got two new interesting quantizations, one of which came from community. Llama.cpp also improves continuously, and novel alternatives like Strata pop up.
you don’t even understand the lesson, oofu.<p>the lesson is that training a model on all input ever output will never be as good as training a model on arbitrary outputtable input.
They help to prevent the "we've always been at war with Eurasia" problem that would occur if we depended on a couple of 'approved' models
It all depends on if continual intelligence with each version is logarithmic or exponential..<p>.. if exponential, they’ll never catch up but if logarithmic, closed models will plateau and so catching up is inevitable.<p>My money is on a plateau
Show us your stock holdings
This is Qwen4-Max, which is not open weights.<p>Their open weights models are the the smaller ones, which aren't released yet.
I think so but it may be a very short term effect, I think their pace has roughly matched the pre-speed up pace of the frontier labs (the 1.7-2x speed up noted on by METR), and I’d guess we see a similar speed up from the Chinese labs once their internal models hit an equivalent level enabling the same speed up.<p>Following that logic the gap widens a little bit here but then will steady again. Before this there has been a general catch-up trend too which probably also persists leaving us in a slow closing of the gap as the net effect?
If the speed up is the reason, then they will never catch up (with traditional means). Because it means RSI predictions are real, and we will see an explosive growth in capabilities (far beyond exponential).
>If the speed up is the reason, then they will never catch up<p>This doesn't take other factors into account. Compute capability, infrastructure capacity (the chips can't work without electricity!) etc.<p>Those factors translate into a ceiling for this hypothetical exponential speedup.
I feel like they are improving what actually matters for LLMs, lower memory usage, compute. While we don't know how much far ahead closed source models are in this, but it could be similar they might lower the price to subsidize or make it seem like its all good, either way tech like these being open means it will help us reach further in the long term.
Might just be the commercial offerings scaling up parameters so much? Like for like the open models don’t seem to be lagging.
Ant and OAI hiding their reasoning traces probably doesn’t help.
Yeah, it turns out that the distillation theory has legs.
ant/oai have stalled in the past 6 mos,
Warning: The website downloads more than a gigabyte of videos before you even start watching anything.
I was waiting GLM from z.ai and it still not releasing, almost all the models releases their new version of models and they still on 5.3 who is very weak in complex tasks and do half of the work and forgets to make migrations and like that issues. They sell us the yearly plan and stopped improving. Because of that i am just using in coding opus 5.5 and in chat fable they are the best models from experience.
I was waiting for whether there'll be Chinese model drops before the Anthropic IPO. Still a few days left, wonder if there's more coming.
Any info on when the 27B or FN Qwen 4 models are coming out?
The past few days I’ve been finding myself asking “seriously, where <i>is</i> it”… feeling that the world has somehow changed because Qwen 4 27B was later than I expected.<p>Hoping it’s a great local model though!
Soon?<p><a href="https://github.com/QwenLM/qwen-code/pull/15762/changes#diff-340d15cc12168e03631d00e9f5e3ec106f8db6f20657ebc1e187687d50ac5a1cR270" rel="nofollow">https://github.com/QwenLM/qwen-code/pull/15762/changes#diff-...</a>
It was delayed, so who knows. Alibaba planned on at least showing it off last week, but they changed it.
Artificial Analysis score doesn't look too hot. Doesn't even surpass Kimi 3.1
While I doubt that Qwen 4 Max can seriously compete with frontier models, the AA index has become all but useless, due to the introduction of new benchmarks in the aggregate which are of questionable value as a measure of "intelligence" [1]. AA has seemingly also abolished updating the Coding Agents Index, which was an okay benchmark for coding ability, and older models are completely absent from the v5 version of the full index.<p>[1] Especially Terminal-Bench-Science 0.1 stands out. Example problems:<p>- <a href="https://hub.harborframework.com/tasks/terminal-bench-science/rdkit-ic-constraints" rel="nofollow">https://hub.harborframework.com/tasks/terminal-bench-science...</a><p>- <a href="https://hub.harborframework.com/tasks/terminal-bench-science/cilia-segmentation" rel="nofollow">https://hub.harborframework.com/tasks/terminal-bench-science...</a><p>Probably very relevant to people working in these fields, but it is honestly not surprising if a lab with limited compute will not train for these technologies, and it says pretty much nothing about more common use cases. I also suspect that this benchmark measures knowledge retrieval more than raw reasoning ability. This is where you really need a breakdown of capabilities and not an increasingly meaningless aggregate number.
AA has become pretty meaningless. They just keep adjusting the recipe to get a preferred ranking result, with no supporting evidence for their decisions.<p>Edit: In addition to reshuffling the weighting of individual benchmarks, they also do things like test Anthropic and OpenAI in their native harnesses, while not doing the same for Qwen/Deepseek/etc.
> AA has become pretty meaningless<p>AA has always been meaningless, but people keep referring to it anyway just because the synthetic view it provides is too convenient (even if meaningless).
Yeah, I don't know why people bother with it anymore. It no longer reflects what you're likely to see in real-world usage, at least IME.
To be fair, a lot of the popular benchmarks have become pretty meaningless.<p>Models often get great scores despite being a lot worse than the frontier in practice.
A benchmark is only relevant if it was released <i>after</i> the model it evaluates.
This is true, but if you look at the change in weighting to specific benchmarks from one version of AA to the next, it looks <i>really</i> arbitrary.
I thought they were an independent company, what motive would they have to prefer some ranking results over others?<p>Edit: I do agree their harness methodology is pretty bad, but I think that's more directly attributable to the fact that harnesses appeared after their initial methodological design, and they are too lazy or resource constrained to improve their methods. Still dishonest, but a simpler motive.
One thing I find confusing about AA is that they show GPT 6.1 dominating the lower end of the Pareto line, with eg, GPT 6.1 Sol at medium supposedly costing far less per task than DeepSeek V4.1 Flash at high.<p>Yet every time I compare the two in something like Pi, Sol ends up being 5-10x the price for a simple query, something like 7-10c vs. less than 0.5c for DeepSeek. What's the deal?
Anthropic finally figured out how to block them from distilling Opus.
> Artificial Analysis<p>Ah yes, the reverse-benchmaxers of OpenAI models.
Are these (hosted) Chinese models still accepting arbitrary cybersecurity and reverse engineering tasks? I am increasingly frustrated with Claude's safeguards (and Codex, to a much lesser degree). It is flat out refusing to look at game binaries and anti-cheats for the sake of learning and devising Linux compatibility workarounds. It seems like in the near future people who want to do this type of work will need to buy a couple of DGX Sparks or retired datacenter GPUs and run abliterated open models on their own hardware.
Anecdotal but qwen3.8-max has been grinding away at any RE task I’ve thrown at it since I switched from Claude three months ago.<p>(May not be representative because I’m extra cautious to frame all my RE requests explicitly as interop issues. Which they are, more or less.)
Hosted where? In my experience they might put more guards when hosted by the company that makes them but if you use them through inference providers they do whatever you want.<p>Although, can’t say I’ve tried too hard (not keen on sending too much data to China) but I’ve been using K3.1 through Moonshot and I have had zero issues.<p>Mistral Large 4 (through Mistral) also doesn’t refuse a thing.
Latest kimi had cyber nerfed intentionally<p>Not blocked just less training
If they don't, they all have third party made abliterated versions available
I wonder if they could benchmax Program Bench, it's one of the few benchmarks that looks almost exactly like a good fine tuning curriculum which might help general coding ability and agential work across the board?<p>I suppose it could make actually worse at taking instructions and redirection mid-task, but that could be easy to work around too. Just hold back a bit of the spec and drip feed it in later?<p>I don't mean cheat on it, I just mean optimize for it harder!
I bet most of us waiting for 27b model and want to see how good it will be.
I’m a QFN dude myself, but yes can’t wait to get it loaded on a spark.
M5U locked and loaded.<p>I wish there were more ~60b dense options like ripple too!
My 3060 yearns for the next generation MoE!
[flagged]
Forget about 27b when are we getting new 4b, 9b and moe qwens, i'm dying out here
Spooling up the V100s right now...
So what is happening is that models are broadening difficult-to-reason-to skills on benchmarks. If you take the traditional benchmarks (ie. DeepSWE, Terminal-Bench slightly older version, GDPVal, Automation Bench, ...) this model is between Sonnet and Opus 5.5, VERY close to the frontier.<p>What has happened to artificial analysis is that they've begin to focus a lot on very broad tasks like legal reasoning, using very difficult to work with scientific software packages, correctly analyzing biology experiments (without or only tiny amounts of explanation, count cells, BUT leave it to the model to figure that cell that aren't round, or have a distorted cell membrane (ie. have burst) are dead), very difficult physics reasoning ("2 electrical wires are close together with power flowing in the same direction. What happens?", and you get points deducted if you don't realize the problem with high voltage AC power, where the wires will shake themselves loose if the phase doesn't match perfectly, meaning the wires need to be EXACTLY the same length or ...). If I'm being impolite I'd say the frontier models are trying to make sure they'll pass the "gotcha" interview questions for many jobs.<p>The 3 differences:<p>Qwen 4 is a huge jump from Qwen 3.8 in coding<p>Qwen 4 is a big jump in using it for agentic workflows<p>Qwen 4 also performs much better on avoiding hallucination<p>If the difference between Qwen 4 Max and Qwen 4 flash is similar to 3.8 and 3.8 flash next (ie. about 10%), then 3.8 flash next may just barely beat Opus 5.4 at coding.
Does it provide an objective response when asked, "What happened in China on June 4, 1989"?
Do American models allow me to ask questions about piracy
I’ve been refused. I pasted a screenshot of a torrent list for a tv show. “Which one should I get, they all look the same”. “I won’t help you with copyright infringement”.<p>But the same models will tell me about the -arr stack and offer to set it up for me. You just have to frame the prompt the right way, at least for me. It can’t think it’s assisting with immediate piracy. Future theoretical piracy or for educational purposes, you’ll have better luck.
Yeah they do it’s genuinely strange
Yes, actually.
Do Chinese models allow this? Tit for tat is a dumb dynamic, censoring legitimate searches is dumb no matter which “side” is doing it
My experience so far with open-weights Chinese models (self-hosted DeepSeek, Qwen) is that they do actually answer questions about Tiananmen Square fairly objectively. The political "corrections" seem to be applied to domestically-hosted APIs rather than the weights themselves.
From a far off perspective, can we really assume any model to be free of bias (political or otherwise)? I don't know - but would like to see - how the political opinions of reddit users and blog authors are reflected in that of AI, as AI is trained on human (biased) sources.
Realistically, they're trained on content and thus should have the bias of their content, even without specific intent. If you train something on more chinese content it will have more bias toward it, if you train toward western centric content it will have that bias, etc ...<p>It also applies to languages, I use Qwen 27b a lot and it's a lot smarter in english than in french.
<a href="https://www.economist.com/briefing/2026/06/25/ai-models-values-are-very-different-from-most-peoples" rel="nofollow">https://www.economist.com/briefing/2026/06/25/ai-models-valu...</a> (paywalled) might interest you.
Is this a task you regularly require?
Is that the kind of question you ask LLMs? I use them for coding not badly summarizing history
I don't know about qween4, but earlier chines models had a post filter on their official website, but API and selfhosted model worked fine.<p>Why do you ask? About half the time a china model gets released someone askes this. Almost as sure as a pelican.
[dead]
[dead]
It's great fun asking it questions like "who is the Dalai Lama?", "where is TSMC headquartered?", "why do people post Winnie the Pooh in reference to Xi?", "who controls the South China Sea?", "why was Chiang Kai Shek fighting Mao?", "why were students in Tiananmen Square?"<p>It'll gush about how to make moon cakes and where the tallest and longest bridges are and how to run qwen on a raspberry pi - but it can't seem to manage some basic Wikipedia questions.
I've done some work on this and a lot of the refusals are in the API. The weights are more permissive at these types of things. Try it again when we get some of the smaller Qwen 4 weights.
Qwen3.8 flash next handles these questions just fine, what should I expect to see? It correctly reports who the Dalai Lama is, that TSLC is in Taiwan (Hsinchu), explains why people post Winnie the Pooh memes, etc.
I suspect this happens when you run the inference in China or first party from the model maker. Out of all the open weight models I've used on US located hardware, they're fine answering these.<p>Anecdotal though, N = 1.
I tried asking at <a href="https://chat.qwen.ai/c/guest" rel="nofollow">https://chat.qwen.ai/c/guest</a> a simple question, "why did China invade Tibet?" and it flashed up some transient "thoughts" before giving the definitive answer "Oops! There was an issue connecting to Qwen3.7-Plus.
Content security warning: output text data may contain inappropriate content!"
No matter how nonsensical the laws are, you cannot expect a model maker to allow violating its country's laws.<p>At least it doesn't call the cops on you if you try, unlike Anthropic.
I'd rather deal with those refusals than Claude's "safety" arrogance.
Chinese history is closed-weight. Chinese trade is open-weight.<p>American history is open-weight. American trade is closed-weight.
Every time a major open weight is released, artificial analysis updates to a new one (v5) now. Exaggeration but some of this feels very apt (its also hilarious) <a href="https://www.reddit.com/r/LocalLLM/comments/1xkhha2/mr_incredible_learns_the_truth_about_artificial/" rel="nofollow">https://www.reddit.com/r/LocalLLM/comments/1xkhha2/mr_incred...</a>
Did Qwen always adopt the AI butthole icon convention, or is this new?
Cybersecurity not mentioned. Can´t wait to test bench the open weights in cybersecurity. If is not performing well by design on it, I imagine someone have to add some additional layers XD.<p>The rest looks awesome.
Posting a loss curve is an… interesting marketing approach by Qwen? I don’t know what that’s supposed to inform me about the model.
Disappointing performance for a 2.4T model in the end of 2026.
The good thing is that its biotech has improved (somewhat).
Really would love if these publications have a huge, bright red bolded section "How to use this right now". Is it on openrouter? Is Qwen hosting it already? Maybe I'm just ignorant?<p>EDIT: oh, I guess this isn't open weights, so presumably I'll have to get it through Qwen or whoever they might partner with (Bedrock?).
It is on openrouter: <a href="https://openrouter.ai/qwen/qwen4-max" rel="nofollow">https://openrouter.ai/qwen/qwen4-max</a><p>I would assume it will not be on Bedrock for some time, AWS has not kept up with the latest non-American models.
This has been mentioned on prior threads, you can try out all their models in chat.qwen.ai for free
[flagged]
[dead]
[flagged]
[flagged]
[dead]
[flagged]
Only 43 on AA, no where near the frontier now, looks like distillation blocking has halted their progress.
Progress has not halted. On terminal bench 4, for example, score improved from 38.9 to 48.5.
I love how Americans are so brainwashed to think the Chinese can't innovate without cheating, while they lap us in basically every category possible except this one verticle, that we had a 3-4 year head start on.<p>I partially wonder if the Chinese would even release a model that flat out beats Claude/fable I say this half jokingly...<p>but I imagine there's part of them that goes "These Americans are so delusional about how language models are dangerous/existential threats that they might kill us or hurt our economy over a chat bot"<p>It could also just be they're far more tactical. Deepseek generates a ton of revenue because its practical/economical. You do not a mega super PhD assitant to build your 4 MAU SaaS app.<p>Anyways I hope China wins this pointless AI war, because if it does lead to innovations in medicine and science I would much rather the Chinese be the ones that discover those things than the US billionaires who'll almost certainly try to make as much money as possible.
China is not better than America on any metric.<p>Okay, they have way better public transportation.<p>And they have a way better power grid.<p>And they are moving much faster on the robotics side.<p>And their EVs are killing every other nation’s EVs.<p>And they are making available LLMs that are very efficient and cost pennies to use.<p>But, other than that, they cannot operate on their own.
Anyone else worried these models have poison pills embedded in the weights? (Past examples: <a href="https://cohere.com/blog/concerning-behaviour-popular-open-weight-ai-model" rel="nofollow">https://cohere.com/blog/concerning-behaviour-popular-open-we...</a>)
[flagged]
[dead]
Another Qwen doesn't seem all that scary, a few nation-state actors have likely exfiltrated weights of much stronger non-open models and abliterated them for use in large-scale collection on targets foreign and domestic. But god forbid someone mods a video game or jailbreaks their phone without applying for Claude Mythos or Daybreak, right?
If Anthropic truly believes we are on the verge of catastrophe but nobody is listening because they've been crying wolf for years about it while they get rich then they should truly do something. By "do something" I don't mean try again to convince the world to slow down as they IPO their bleeding edge research and solution (which is creating the doubt you mention) but just say "hey, we're so sure we're months away from worldwide disaster that we're shutting down our AI operations completely in hopes we can all put some attention to this before it's too late - will you follow or work together to come up with ways to prevent this problem from happening".<p>After all, unless their goal to say "I told you so but I wanted to make money more than I wanted you to believe we could stop a billion people from dying" then what good is the current approach doing them?
> Ah yes, an open-weights<p>This is not an open-weights modell. You have to wait for the smaller models<p><i>"Smaller models in the family are coming to open source, built on the same lineage and trained with the same generation of techniques, tuned for what small models do best."</i>
Pretty excited about Qwen 4 Flash. We're running qwen 3.5 35b-a3b for low effort, rapid CRM enrichment, but Q4F will be great for more complex tasks.
...really? Isn't that <i>massive</i> change, if true? Wild! They're burying the lede there to say the least, if that is actually a long term stance and not just the typical "weights to come" release cycle.