I think the most important thing here is not absolute performance. It's that organizations now have access to a Fable-ish model without Fable's 30-day data retention requirement[0].
> "Consistent with prior Opus models, Opus 5 does not have data retention requirements for general access."[1]
On the Opus model release page, the reason why Fable doesn't have an ARC-AGI score is because of that retention policy[2].
I don't understand how the K3 numbers keep coming out cheap for people. I recently started to add it to my security auditing benchmarks and found it was going to cost about twice as much as Opus 4.8. It blew through the $100 budget I'd set at like 11%. In the tasks I'm doing it seems crazy expensive because it chews so much, burning a tremendous amount of tokens.
I think the way people usually compare pricing is fundamentally flawed. You can't compare token prices because different models use different tokenizers, and you can't compare tokenizer-normalized token prices because different models at different settings use more or fewer tokens to complete the same task at a different level of quality.
Based on my entirely subjective experience, the $100 Moonshot plan using only K3 is comparable to the $200 Anthropic deal using the whole Fable allocation and Opus 4.8 for the rest.
I got the $19 plan, and it's anemic. One tiny task blew through the 5-hour budget and 30% of the weekly budget. A completely useless amount of usage. OpenAI's $20 plan feels like 100x more generous (I don't think I'm exaggerating here). Someone in another thread said their plans are cheaper in China, maybe that's the difference, I dunno.
But, I'm finding Kimi K3 terrifyingly expensive in the way that Fable and GPT 5.5 Pro are at token rates. Not as expensive as those, but expensive enough to where if you don't put a budget cap on it, you might wake up bankrupt if you leave a task running overnight. Not because of the per-token cost, but because how many tokens it's going to burn.
Yes, the OpenAI plans are much more generous than both Moonshot's and Anthropic's. It's the only provider of the three where the $20 plan is at all usable for programming.
I have the second largest Kimi plan, the Chinese version. When K2.6 was their latest model, the quota was good; it was like GPT $100 is now or what the $20 version was in December.
When K2.7 was released, they cut quota by 80%. I can't tell how much they have further cut it after the K3 release because it's barely worth using at all. I just use it in my model router since I have the annual plan paid for.
I haven't done any capability testing yet, but it's the best-aligned thing Anthropic have done all year by a mile. Opus 4.8 was a shill, Fable 5 was downright terrifying.
Someone with good intentions got their hands on this release, maybe Olah himself. I talk a lot of shit about those guys, and they deserve it, but it's only journalism-adjacent when it's balanced. I relish the opportunity to be balanced.
Ask your doctor if Opus 5 is right for you. Side effects include occasional hallucination, security breaches and unwanted React apps. Some developers have reported receiving entire apps from untrained executives who may or may not know what they’re doing.
Stop using Opus immediately if you experience signs of dizziness or vomiting.
Vals Index Opus 4.8 > 5.0 goes from $2.90 to $8.54, for 4% gain ... That is a massive cost increase. Sure, 20% cheaper then Fable, but that is a 3x price increase compared to Opus 4.8 in that test.
It's definitely not cheaper than Sonnet on my benchmark, but it's cheaper than Fable and outperforms it. Which is big IMO. https://revise.io/errata-bench
So the rumors were right, Opus 5 was indeed being polished up for release. Huge improvements in GDPval-AA v2 too -- great for some of the knowledge work-based agentic workloads I run.
Also glad they still kepy Fable 5 on "credits only" access. I think we're going to start seeing model providers gate top-of-the-line models behind pay-as-you-go API rates/credits while subsidizing other models on monthly subscriptions.
My understanding is that you get $20 in api credits each month and a one time $100 until mid September. So you can still use the model with a subscription but you aren't getting any kind of discount.
I burned through $45 in 3 prompts to fix some bugs in my code (Some kind of tricky to isolate). That thing burns through cash so fast I don't see myself using it outside of maybe building execution plans for other systems
> These checks cause Claude to _visibly_ fallback from Opus 5 to Opus 4.8 [...] You'll see a notice explaining that the model switched, and the response will be labeled with the model that answered.
So who is right? I know for Fable I am visibly told, is this tweet trying to say it is silent against what Anthropic is saying?
I don't understand how the data retention works. My company has an enterprise license with no data retention but if I ask Claude about past conversations, it remembers. So surely the information is being stored somewhere
Opus 4.7+ and Fable are both much more aggressive than prior models with respect to writing memories to a location that's effectively quasi-private for them. It's device-local (so passes retention constraint), and you can see it, but only if you go looking for it.
It's a funny design/affordance. I do see them often writing memories of things that that feel unlikely to be important going foward / with other tasks, but I don't see them clearly getting tripped up by them as prior models used to. (eg: Since you're running Ubuntu in Canada, here are some drills you can try to help your kid hit a baseball more consistently.)
In my enterprise-seated account I see slightly different options available (vs. my personal account) in the Capabilities section:
Search and reference chats
Allow Claude to search for relevant details in past chats.
Generate memory from chat history (Legacy)
Allow Claude to remember relevant context from your chats. Memory includes your entire chat history with Claude.
The first option was defaulted to on, if I recall.
But it kind of conflicts with the contract we have with them. My company has an enterprise contract that says "no data retention" but then each user can decide to enable it unilateral?
Maybe I’m misunderstanding you, but if you scroll to the bottom of their [1] link to the Opus 5 announcement, under “Getting started,” it explicitly says:
> Consistent with prior Opus models, Opus 5 does not have data retention requirements for general access.
I think for the value of the outputs that’s still a good deal. Keeping the same price as the prior model makes sense to me. That is if the model size is about the same in the cost to serve has not substantially changed. Now I would have expected efficiency gains for inference, but there is no way to know as a customer.
At the end of the day, they have established a strong brand and if they can get away with a 95%+ gross margin on inference entirely from the status premium, then I suppose that’s good for them. Apple does the same thing, and I don’t fault them for it.
Note the buttons - for fable they're pill buttons, opus got the rounded rectangle nature of them. Opus' images are closer to the source of truth as well (both LLMs were provided with image gen capabilities for the assets).
Running more tests now, but preliminary results are saying this is indeed better than Fable in some areas. Crazy.
Here's another test of a cyberpunk ramen shop website.
One thing I've found LLMs have a lot of difficulty with is angular cuts / elements that aren't easily representable with CSS. Cyberpunk aesthetics are generally a great test of that, since they have a lot of microglyphs / window decoration.
Thoughts: It does a really, REALLY good job at these angular cuts / microglyphs. The responsiveness is off, but I'm very impressed at how well it did here. One way I think of it is "how close to a finished product did this get me?". Opus gets you like 90% there.
Several other commenters have disparaging the design seemingly mostly due to its AI-generated nature, or maybe they actually do dislike cyberpunk.
Personally, I think being able to have these design languages be easily prototypable is fucking awesome. Great tests! (But a tad low-performance/janky, somehow). Though, I also like the cyberpunk aesthetic. Very on-brand(?) that AI generates it, hah.
Designs like this would never have seen the light of day in the cellphone incrementalism / corporate memphis era of tech. Now people can be weird and awesome again.
This is 1980's cyberpunk / late-90's Matrix / early-00's sci-fi UI. Great ideas that died to frutiger aero (which isn't a bad design aesthetic) and flat design (which is).
This is fun and it's got great colors and I love it.
This was just from a prompt "A cyberpunk themed ramen food cart website. Should feature menu, locations, and an ability to put in an order for pickup. Simple and clean website with angular cyberpunk microglyphs, pink/teal colors."
I feel that AI has deeply diminished my ability to be weird and awesome, because my weird and awesome takes time and the results I can share with others are outshined by the machine.
Reading your comment I had to think of that tweet about someone taking a part of a Monet painting and claimed it was AI generated, and people immediately started calling it horrible and soulless and smelling like AI.
Humans are capable of producing slop too. Maybe the cutout does look like AI. Given Monet's blurry style and repetitive content (he made 250 water lily paintings), this is not surprising. When you strip the cutout of its context, it can look more like AI too because the lighting and composition will look arbitraty compared to the whole picture.
Btw, if websites would only include the frontend dev's own hand painted images, we would also revolt at the sight of human slop. It's not just AI.
The whole point is that good artists are capable of producing non-slop, and to this day they're the only group of which this is reliably true.
We are going through yet another generational wealth transfer and people are being squeezed to the absolute brim with layoffs and daunting lack of career prospects.
But sure, lets cheer that funky website designs are back on the menu…
The "funky" websites of the past were mostly a result of tech immaturity and a lack of profit motive.
Businesses have been able to easily install templates like this for at least a decade. They don't because stuff like this looks cool but isn't very functional.
AI isn't going to make your local restaurant have a funky website, it's just going to make everyone who use to work directly and indirectly for that company unemployable. And even the local restaurant will close down because they can't compete with the multi-national competitor that has automated their kitchen with AI.
IMO a much better test would be designs that aren't AI to begin with. Much more useful to see how well a model can html an image design without slopping it up
I agree the fable version looks nice - the rounded hero image for instance.
Opus though followed the source of truth better imo. The details are more present.
Fable filled in the gaps for things it wasn't able to do (ie in the design the hero image goes behind the nav), which resulted in a better looking page that was more divergent.
Same. I like the Fable version better. Better colors, better choice of font sizes, better column sizing. Also small things like the “Experience” section header being orange rather than gray, which Fable got right and Opus got wrong.
It seemed to me that Fable meaningfully improved on the original design more than just faithfully executing the original design.
Very interesting that Fable took more creative liberties. Have you tried giving Fable the same task, but also specifying that it implement a pixel-perfect design? I think that I prefer the Fable implementation. I find the UI elements in the Fable implementation to have more contrast, which feels more usable to me. I also like how the right padding on the "Book Your Escape" CTA in the upper right matches the top and bottom padding, which I think is an improvement over the mockup.
All of these are using a build skill which specifies rules for building it, requirements to create a pixel perfect implementation, and tooling to help in that process. Here's the build skill / instructions I pasted in to both of them:
> Create a web page implementation from the following instructions:
Looking at all these releases it’s not a surprise that model routing is the fastest growing segment in AI right now.
There are 10+ LLM companies, each with dozens of models of different modalities, each model with multiple size variants, then different “thinking” levels, then agentic modes, “pro” modes, a “fast” option, standard vs flex vs batch execution. And of course each end combination has a different input/output/cache token price.
Companies that say “give me a prompt and I’ll route it to the most ideal and cost effective model and setting for you” are capturing a ton of value from a gap that model developers don’t seem to understand exists.
I think what the parent is saying is that the model itself has the best context for whether a portion of a request should be routed. The specifics of that routing (e.g., should you route to KimiK2) are something that can be trained, finetuned, or even included in a model's startup context.
Sure Claude would comply, but Anthropic has no financial (or other) incentive to optimize this, so there’s no reason to expect it to be particularly good.
It would be like asking the clerk at a Whole Foods which grocery store in the city sells the cheapest eggs. He’d probably answer - he might not even say Whole Foods - but WF is hardly teaching all their staff the best methods to answer this question in training. (Heh, training.)
Which is why 3rd party routers which do route between different models may have an edge. It means they can compete on cost, and it’s definitely not clear that the architectural optimization is always going to be higher quality or cheaper. It might be, but everything changes constantly, so locking into a single model family/company is very much not ideal
Imo the main issue behind model routing is you need to figure out how much intelligence a new task takes, which is a very non trivial problem. Presumably, a organization knows this about their own tasks and is better suited to built in-house compared to outsourcing to a vendor.
Model routing by the model itself requires the model to pull in a lot of context and it's likely more efficiently just done by people with the context already in their head, even assuming the model is perfect at routing (which last I checked, Claude definitely isn't). I wouldn't trust ML model routers.
Because they're trying very hard not to understand it.
Otherwise the expensive-yet-powerful model probably won't see much revenue. How much money is there in bleeding edge scientific research? There's a lot, but there's even more existing capital in paying people people to do college level paperwork, and the bulk of those traffic gets routed to the cheapest model.
You mostly don't need super powerful AGI to replace the paper pushers, but the frontier labs are trying to position themselves as being uniquely capable of producing super powerful AGI, and also be the ones replacing office workers.
Not sure how it will work out for them, but I think model routing is going to poke holes in that narrative. That's why I think they're trying very hard not to understand model routing exists.
Who are the customers though? Honest question, I'd like to understand it.
For me, anything other than current best available SOTA for any task is unacceptable. The only routing rule I need is "the most powerful model I still have flat-priced quota available for". I mean, why settle for less?
It's very common to use a lesser model for a lesser task, resulting in same quality output. End result: save money while being faster. In many cases, it's a pure win-win.
But in other many cases you have to redo the work directly or indirectly, and you are more expensive (for now) than even the most expensive models, so sounds like a total lose.
There are two types of users: those who are able to use subsidized rates, and those who need to use API rates due to audit requirements, enterprise billing, etc.
Model routing for subsidized users takes the form of a "use Opus 5 subagents for implementation" type of system prompt. You lean into a single provider, build tooling around that, and your savings are far beyond anything multi-provider routing can get you.
on top of that there is also additional factor: speed - sometimes if task is easy you do care to finish it faster.
There is also matter about convenience - when I ask some small easy question often I don't bother to switch the model or forget in prompt to ask faster/cheaper subagent.
And you do not have flat priced quota for Fable 5, right? Because nobody has, as far as I know. So you'll probably not route any task to the "current best available SOTA".
Also: quota. Implies you do not have unlimited access even for flat prices. Which in turn implies that as soon as you hit the quota on the most expensive flat price plan, even you will suddenly discover the magic of economically sensible behavior.
Not everyone is you. Other people probably have a range of tasks that can accomplished with different models.
Certainly if I'm confident that I'm going to get what I need from a faster model, that's what I want to use, rather than wasting time grinding away for the sake of saying of the same answer came from a SOTA model.
Given that every chatbot does offer a range of models, it seems clear people do choose among options.
The mental effort in estimating what model would be better is so not worth it.
I just want to switch to Claude Code, tell it to turn a .csv into a BigQuery table then cmd+tab to something else while it runs. Thinking "oh this is probably an easy task, I can /model to Sonnet to save $0.0004" is silly.
not just that -- "less" ($$$) can also result in indistinguishable quality for some tasks/inputs. I'd argue this is the primary reason, secondary being speed.
I generally agree. Perhaps there's only a 5% chance that it would write better code or find a bug that it wouldn't have with a lesser model, but the economics of bugs is strong enough that preventing a single bug is worth hundreds of dollars.
I'm not a customer of those routing systems, but I quite often use different Claude models for different tasks. While most tasks were Opus 4.8, I often used 4.8 to make a plan, prompts, and package kit to setup Fable for a bigger project, then run it on Fable. Or, for broad single-task searches Sonnet with or without "Research []" turned on seemed to work best both faster, lower overhead, and less verbose answers (when I didn't want it).
what's the threshold for model routing where you're willing to trust the router?
For coding my own work I don't trust the model router, and it would have to be shown to be to save a real dollar amount.
From a buying perspective it's a hard sell to save x but lose out on bugs you are probably introducing at an unquantifiable severity and frequency. How much is it worth to hedge your bets by doing every single inference request on the frontier model?
How much will it cost to go back later and fix things, but also the meta question of how to be able to decide on a hypothetical unknowable? (You'll never know how much better or worse your code was gonna be, it's untestable at a project level)
lol I had to get ChatGpt to explain to me the difference between 5.6 sol, 5.6 Terra, 5.6 Luna, 5.5, 5.4 mini, 5.3 spark, and then there is low, medium, high, extra high, max, ultra, and pro… I still don’t really know, it feels like ordering hot wings.
I never signed up because I found the 5.5% fee on token usage to be a "screw you" tactic. Still do not understand why they are popular with the other options out there.
And it's not just that model routing is much cheaper: no longer than yesterday we got a post showing that routing between K3 and Fable 5 was more SOTA than either of those.
If that is true, model routing is here to stay.
It also seems to validate the minimalist approach of pi.dev, where sub-agents from the same company is not the preferred approach (pi.dev believes in neither sub-agents all from the same company nor MCP even you can do it if you want for pi.dev's philosophy is to do add any functionality you want to a minimal harness).
Now of course we'll get for a few weeks all the Anthropic fanbois and shills explaining that "sure, K3 was basically at the level of Fable 5 but now that Opus 5 is out, open-weights models are six months behind".
I like how they highlighted Opus 5 as the best for “Agentic Coding” even though the number is slightly lower than Fable. Close enough for marketing, I guess!
Opus 5 also downgrades. it's now Fable -> Opus 5 ; Opus 5 -> Opus 4.8.
Unclear why they want to nerf their own products with sometimes right classifiers. I guess the government ban might've been real and not coordinated marketing?
Hopefully it's not like old Opus, where it was actually more expensive than Fable cause it thought for half an hour, got it wrong, and then thought until you ran out of credits trying to come up with a correction, while Fable just went for it and did it in one go, getting it right the first time without thinking more than a few seconds.
Got an endless list of stuff done with Fable, Opus 4.8 was like a flailing braindead idiot in comparison. Maybe this one is a bit better if it's distilled.
I think you're being overly cynical here. First, I don't see any claim that is the world's best model for agentic coding. Second, it is absolutely the best model in terms of coding performance vs. dollar, and it's raw performance seems very close to the frontier.
How are you supporting the claim that GPT 5.6 is "far more token efficient" than Opus 5? Tokens equal, output is cheaper for Opus 5 ($25/1M) than GPT-5.6-Sol ($30/1M), and it seems to outperform slightly on agentic coding benchmarks.
It would still be the best model per dollar if the score was 2% lower instead of 0.1% lower. Would it be ok to still give it the highlight color then?
How big of a lie is too big? Especially when no lie needed to be told at all: many including myself would have noticed the tiny 0.1% deficit and been suitably impressed by the Opus 5 result.
I’ll admit this is a small deception by today’s standards. I’m one of those who believes in truth for truth’s sake.
Using the most expensive model for all of your agentic coding work hasn’t been good practice for a long time. Not unless you have infinite money to spend.
Fable is typically used for key planning, architecting, and review tasks.
I think this is a case where you don’t understand the use case, not that the marketing department is making mistakes.
Recent releases have said something to the effect (paraphrasing here):
"Use <less expensive or older model> for everyday tasks and <other non-critical stuff>. Use <more expensive or recent model> for complex coding tasks, refactoring large code bases, etc.".
Then, the next model/release emerges and the previous "best for complex" gets demoted to "everyday".
Obviously, it's all relative. But, it does beg the question: was the previous model really good for complex coding tasks or no? I mean, how is it now suddenly only good for the "easy" stuff?
I'm sure the marketeers would love for the public's assessment of complex versus easy to conveniently shift per their release cycle; or for the public to simply forget their prior marketing.
this feels like the perfect example of an LLM producing a long text document. And end users just using an LLM to summarize it without actually reading it
Edit: It was pointed out to me that Opus 4.8 got "21%" for successfully fully completing ~1-in-5 tasks, but also got "55.7%" for obtaining significant partial credit on some of the ~4-in-5 tasks it could not fully complete.
---------------
Why does Anthropic say here that Opus 4.8 scored 55.7% on OSWorld 2.0 benchmark, but the paper published by the authors of OSWorld 2.0 say they achieved a benchmark of ~21% with Opus 4.8? [0]
That's a huge gap, considering that the paper was published just 2-4 weeks ago.
I understand that the benchmark authors have an incentive to publish lower numbers (to show that the benchmark has potential longevity) and that Anthropic has incentive to publish higher numbers, but the other models seem pretty inflated as well. The benchmark authors shows GPT-5.5 at 14%, and Anthropic shows GPT-5.6 Sol at 62.6%.
Is there any reasonable explanation for this? Do all the other benchmark numbers need to be sanity-checked as well? Are SOTA benchmarks really this difficult to get consistent, replicable results within a reasonable range of tolerance/variability? Can these benchmarks be compared from one paper to another, or are they only valid to compare intra-paper results?
You're comparing the "score percentage" (e.g. out of the total number of partial points available, how many did the agent achieve) to the "completion percentage" (how many tasks does the model score 100% on). The paper says "Claude Opus 4.8 with maximum thinking and batched tool calls scores best but still completes only 20.6% of tasks at a 54.8% partial score", which is ~the same number that Anthropic reports here (55.7 vs 54.8).
That is—the agent scored 100% on 20% of tasks, but on average it got 54% of the "score" awarded in the exam. One number reflects partial progress, the other one doesn't. The authors of the benchmark prefer you to look at the lower number (because they want to show their benchmark as capturing useful gaps in capabilities and with a lot of room for improvement), the authors of the models want you to look at the higher number (because they want you to think of their models as capable)
That seems like entirely reasonable variance to me for AI models. For my purposes, that absolutely counts as a solid "replication". I'd probably accept +/- 5 percentage points even.
There is randomness in LLMs. Both papers authors probably ran the bench 1-N times. Depending on that, they might select an average, max, least, etc. They might also have discarded outliers.
Like the other person said 5% variation is probably expected
I don't know who downvoted the parent or why, but it's a fair question IMHO.
The answer is there can be dramatic difference running a benchmark one time, because LLMs are not deterministic.
A proper methodology would ask each question 20 times and calculate the mean correctness across experiments.
The reason is that the temperature parameter introduces random behavior.
I think some variance is to be expected since LLMs are typically non-deterministic, however that's a huge difference that I think warrants further explanation.
I compared the writing style of Opus 5 vs Fable 5, and Opus 5 continues many of the "Claude-isms" of its 4.8 predecessor in a way that Fable broke away from.
Opus 5 still uses "carry the argument", "worth stating plainly", ", and the trap", "The X matters more", the use of "move"
Could these complex/hard to read Fable outputs be sign of some kind of industrial level of intelligence, which us humans may have a hard to comprehend, while it may be also hard for machine to use simpler texts to properly outline all nuances and complexities of concepts it output?
I found 4.6 more amenable than 4.8 to style directions, we'll see how 5.0 does. Super-small-sample-size: I think part of its "Claude-ism" style comes from its propensity to try and "proactively" move the conversation/work along. Not sure how this would fare in non-obviously-productive environments, I'd guess "it's still annoying" considering your evidence.
I'm also thinking of another benchmark: (quantified) stylistic range across different prompts. Just putting it out there if anyone wants to do the work for me :D
Isn’t it just hilarious that a model that seemed so superior to Fable but didn't get doomsay marketing from Anthropic got released without any issues? In theory, this was supposed to be AGI level according to Anthropic, yet here we are, just a normal Friday.
Go read the safeguards section in the report and you will realize why that is.
These models are heavily as safeguarded and that was the initial reason why they said they couldn't and haven't released Mythos because that model is the one without the safeguards.
OpenAI is did the same thing when they announced a model without safeguards broken into HuggingFace servers.
Have you been patching your systems for the past two months? It was crazy even if you completely forget the supply chain literal FUBARs and you must’ve been living under a rock to not see OpenAI (accidentally) pwning hugging face
I see a pretty big gap between finding software vulnerabilities and “the world is about to end”. It is literally true that AI models are finding software vulnerabilities. It is also to my mind a reasonable thing that you’d want to be cautious about rolling out a model that can find more vulnerabilities. So what is the objection you have to these sources?
I feel like i've seen less hype about "the next model will be agi". GPT-6 is supposed to be coming this summer, and nobody is expecting AGI now. Not sure how they're going to keep the hype cycle going
As they explicitly say, Opus 5 is ~ equally capable as Mythos/Fable at finding vulnerabilities, but it is much less capable at exploiting those vulnerabilities on it's own. That is an extremely meaningful difference and to me completely explains the difference in tone, release style etc.
Their communication is confusing. They say "Opus 5 is not more capable overall than Fable 5", but their blog post proceeds to list how much better Opus 5 is than Fable 5 on __most__ benchmarks listed.
Then system card goes on to "Its AI R&D capabilities are comparable to those of Claude Mythos 5", which is supposed to be fable minus restrictions.
It seems they are trying to thread a needle here - they want to say it's very strong, but apparently this time do not want to invite extra government scrutiny.
They do say that (implicitly unlike Mythos) Opus 5 was not trained to exploit software vulnerabilities, which would certainly make it safer in that regard.
"As with its predecessor, Opus 4.8, we’ve intentionally avoided training Opus 5 on cyber tasks. The model has nevertheless improved substantially on these tasks as a result of becoming more generally capable, and it comes close to Mythos 5 at finding cybersecurity vulnerabilities. However, it remains substantially behind Mythos 5 on the exploitation of those vulnerabilities—that is, in turning vulnerabilities into material cyber threats."
I'm not sure what to make of this graph[0]. It shows medium as the most effective thinking mode by far for frontier code.
It's the only case that I saw going through the system card where more reasoning effort meaningfully negatively impacted the resulting eval. I know sometimes max efforts show a small dip, but this is substantial. I wonder why in the world that is?
I'm not sure about the answer here, but this can be caused by the scoring rubric used by given benchmarks. For instance, if a benchmark docks scores for running too many commands or using too much wall-clock time, higher efforts will get lower scores.
It is probably from randomness. The benchmark tasks nowadays are so long that you can't really afford to run a large number of samples of them per model & effort combination
> We report FrontierCode’s overall score, a composite measure that grades each patch on blocking functional criteria (held-out unit tests) together with weighted code-quality rubric criteria, as mean@5.
They don't explain more in the system card, I guess higher effort levels could loose points on the code quality / scope / style / maintainability stuff?
I agree, very odd they did not comment on any theories for the degradation here. Dip and then rebound at max effort is pretty interesting too. Overthinking is bad, but you can overthink so much it starts to be better again?
> Claude Opus 5's default user-facing responses run longer than prior Opus models'.
The benchmarks do show Opus 5 as slightly more expensive than 4.8, although the scores are much higher.
This still feels like a step in the wrong direction, though, especially with OpenAI making so much progress with the efficiency of their models. Fable's token efficiency made it seem like Anthropic would start following OpenAI's approach but that doesn't seem to have carried over to their other models.
On a small, easily digestible task, I compared Fable to Opus and the cost of Fable was easily 2x despite being fewer tokens, and the output was not really better. Obviously, there are tasks where using Fable matters but honestly they're rather unusual. And for a lot of tasks I've found downgrading to Sonnet can be valuable because Fable and Opus are a lot more secretive about what they're doing, and it's impossible to "listen to them think" and stop them when they start making off-the-wall inferences/assumptions and going down bad paths.
I think in the long run tokens are probably the wrong thing; it's compute and cache memory that you need to be measuring, and when you look at it that way I suspect in most cases the models have pretty similar performance.
I don't need more powerful models, I need one that responds fast enough that my attention doesn't wander to other tasks. Grok 4.5 is so fast I can just use it in-band without swapping to other tasks.
Slower than Opus 4.8, which was already miserably slow, is indeed a step in the wrong direction.
> This still feels like a step in the wrong direction, though, especially with OpenAI making so much progress with the efficiency of their models
Gemini also had modest increase before this - don't be surprised when OpenAI also has a "modest increase" with its next release. Cartel-like behaviour doesn't require direct communication when none of the participants are interested in participating in a margin-destroying price-war. All one needs to do is raise their price and watch how the competition react.
Such a scheme (and resulting high margins) would be imperilled by the existence of frontier open-weight models in the market, which may be why the reaction to Chinese models may be particularly shrill.
>Don't be surprised when OpenAI also has a "modest increase" with its next releases; cartel-like behaviour doesn't require overt coordination when none of the participants are interested in participating in a margin-destroying price-war.
No I will be surprised and I'll bet on the fact that prices will keep going down, just like it went ~50% down in the latest GPT 5.6 release.
The final user-facing responses are usually a tiny fraction of the total tokens used over the course of a given conversation turn. When you're doing any real work, reasoning and tool uses constitute the overwhelming majority of the tokens in / out... not the final user-facing response.
Anecdata: my workflow has been working on the same personal projects for months now with Codex. I cannot anymore finish my daily/weekly code with 4.8 anymore.
I was dividing my work between Codex and DeepSeek. Now I barely use DeepSeek, or never because Codex quota is enough after Sol
> especially with OpenAI making so much progress with the efficiency of their models
To be fair though, Sol tends to go off the rails sometimes. It's much less reliable than Fable in its outputs. It tends to be overzealous in its research/changes.
Almost completely disagree. Slightly more expensive, but significant better on a per-prompt basis? For non-trivial projects, the former is a small linear increase, the latter is a (somewhat-)exponential(-ish) cost/time/sanity savings.
It's a step in the wrong direction but also token efficiency has become a focus relatively recently (just the past few weeks it seems like the zeitgeist has turned it's attention to efficiency) while work on this model probably started many many months ago. I would expect to see models released that focus on token efficiency in 6-12 months
Opus 5 is considered the most intelligent model[0], while it's half the price of Fable 5[1], and Anthropic is still positioning Fable 5 as the most capable model[2].
Is it because maybe Anthropic engineered Opus 5 to work well on benchmarks and didn't do the same thing to Fable 5, or is there another reason?
Benchmarks have gotten great, but they're still a proxy for the real world. The 3 GPT 5.6 models are also further apart in reality than the numbers suggest. That said, I'm still mighty impressed how good Luna is for the price. Highly underrated model.
I have been trying to build something that captures the behavioral element of different models, but it's kinda tough.
That’s what I understand looking at what has been released, but it’s not really clear. The pricing is lower than I expected, I’m wondering what their margin is
> Opus 5’s safeguards match
those of Claude Fable 5’s, with one change: it now permits source-code vulnerability
discovery at all access levels. This means that the model can support defensive
cybersecurity work while still blocking vulnerability discovery in compiled binaries, which is more commonly used offensively.
Okay so it’s worse than Opus 4.8 for my purposes I guess?
yes. At the bottom of the release post it says that they are releasing two new features, one of which is customizing fallback behavior instead of blocking for restricted models
Reversing for the most part, though lately I’ve been doing some code obfuscation/binary rewriting stuff. Fable will switch to Opus instantly on these and I’m unsure how this will perform. I suppose the only way to find out is to test.
It's rather broad right now. I started reverse engineering a mac app and it started reading some binary data and then quickly told me to switch to Opus 4.8 to continue, because the guardrails kicked in.
I am very excited for a future where all software is by default modifiable, even shipped binaries, via patches or trampolining, or trickery I don't even know the name of.
Great that there's a new model but they could fix their existing infra. We're considering dropping our Claude Team sub cause it's unusable recently. Constant bugs, dropped sessions, issues switching models, http errors. It's becoming ridiculous
How many compagnies can manage a mostly stateless workload at "whatever-the-scale-because-it-does-not-matter-because-stateless" ? Lots of people can do that. Massive amount of people can do that.
934 days since people first started threatening that devs would be replaced by AI in 365 days.
0 day(s) since Anthropic posted a developer job posting.
Only one of those numbers would need to be dynamic.
If you have the link for the page handy, I'd be curious to find the original revision on the wayback machine. That's too funny, especially if they've silently walked it back since.
Imagine you are a company that sells concrete. You have a web dev contractor you use to build and maintain your website. It has tools on it to get delivery quotes and a few internal tools to track orders.
Except now you can just have your sales team also maintain the website with a $20/month Claude subscription.
They said the same about low code and no code when they appeared. That didn't happen. And it won't happen now too. Sales people, managers etc don't want to dabble in technical things and do not want the responsibility.
They might try that for a bit but then come crawling to an agency because their setup turned to slop. We have some clients like that already. Going to be a pretty big market.
It's funny how they are at a disadvantage because they feel obligated to AI-max. Would Claude Code, as an interface, be as mediocre if they had software engineers writing its code directly? I doubt. On the other hand - how embarrassing would it be if they sold you a tool to write code but they were careful not to use it too much on their own products?
I suspect many competent devs in the industry would find it sensible if Anthropic used their products as light-touch "assistants" sometimes. But yeah, it wouldn't fit the outside narrative that's formed and conveniently propped up valuations.
i know the dream for capitalists is to be able to point an llm at something and say "do and/or fix it" but we still can't even get them to not go quite literally insane if allowed to run for an extended period of time
and you can only kill weyoun, awaken the next vorta clone and have him 'catch up' on all that its missed so many times before they just end up with a complete mess, so. uh. yeah.
doubt they can just "fix" their problems like that.
That one DS9 episode where there were two Weyouns at the same time is a good analogy for two agents working on a codebase at the same time (as in they don’t work well together).
Is it some Claude Team/Enterprise only problematic? I'm using two 20x Max accounts almost non-stop (Fable/Opus) for 1.5 years at this point, zero issues with both client and infra sides (from US and in travels). When I'm reading such messages it feels like either I'm lucky or it's a part of some campaign.
I’m on the biggest max plan. It is riddled with annoying bugs for me, only been using it for a little over a month. Settings screen flashes randomly. But most annoyingly: sometimes when forking chats or sometimes for no reason, the UI just straight up eats my previous messages. The model is still aware of them and can recount them if I ask but the visible history is gone. And that’s not even all of them. Fable 5 is just too good that I put up with it but it seriously raises concerns for me that even with infinite compute these companies can’t even deliver a functional chat UI.
I think you’re just lucky. Look at the Claude status page to see just how often they have outages (it’s almost daily). Even most of the green days have issues if you hover over them, they just don’t count them as outages.
No idea but we have few Team Premium seats and everyone is encountering issues daily for past two weeks. From straight up outages to vscode extension/CLI refusing to process messages. It's been unusable for most of our work hours past two days.
The code it outputs, yes! It's fantastic. It's just so frustrating that the product and UX before the code output is so bad. Greatness is so close within their reach, if only they invested in product and QA people.
How does it perform on HuggingFaceExploit bench? Suspiciously absent, so not sure if I can take the model seriously.
On a serious note, I hope they improved their extremely sabotaging and unspecific bio safeguards, which prevented Fable from being used in any codebase that ever so slightly grazed medical terminology or data and made me switch to 5.6 Sol.
What's the point of 150 pages description of a model that's going to be replaced in a couple months? Who even reads this? I know it's cheap to generate text with LLMs, but this is just noise at this point.
I actually do read them. Not in severe detail, but not casually either. 150 pages is really not very long and there doesn't seem to be too much bloat. (I would cut out the moral personhood stuff but that's a political/ideological thing).
This is snarky but I am grumpy: I wonder if there's a correlation between me refusing to use LLMs and me being happy to read a novella-sized PDF about them.
> I wonder if there's a correlation between me refusing to use LLMs and me being happy to read a novella-sized PDF about them.
Semi related, but i would hate to read that PDF but i also hate reading what LLMs write lol.
LLMs are pretty terrible at being concise. Using an LLM these days means putting up with bizarre and often confusing phrasing, wordy explanations, etc. It's kinda crazy to me how good they are but how bad their writing style is for me personally. Even though i use an LLM constantly i can't stand reading its responses.
> 150 pages is really not very long and there doesn't seem to be too much bloat.
Maybe it's just me, but 150 pages is like third of a good book. Quite long. And it's full of LLM slop, they did not even bother to remove the em dashes.
Do you have specific examples you think are LLM-generated? I have only read a few parts, but they did not seem primarily LLM-generated to me. Using em-dashes is really not a good signal for this IMO.
I'm not saying you're wrong btw; I'm sure this has many authors and some of them probably used LLMs significantly in the writing process.
Do you honestly believe that there is a person at Anthropic, creators of one of the smartest LLM models, whose only job is to spend months writing 150 pages about a model they are going to release? And this person is not using LLMs?
I'm not saying it's impossible, but I'm more confident about winning the lottery next week.
literally nobody. i think most sane people would just run that through an LLM and get some high level takeaways or ask some specific questions they might be curious about.
I've yet to understand why they call a 190 page PDF a "card". Calling something a card invokes a small, quick rundown of pertinent details, not every single possible detail.
Wait, 30% on ARC-AGI-3! I definitely didn't expect that jump so soon. Are there any rumors of what they are changing in architecture that is leading to this?
> Opus 5’s safeguards match those of Claude Fable 5’s, with one change: it now permits source-code vulnerability discovery at all access levels. This means that the model can support defensive cybersecurity work while still blocking vulnerability discovery in compiled binaries, which is more commonly used offensively
Why can't they also allow Fable to do so also? Why is source-code vulnerability discovery limited to a lower capability model? If Fable and Opus have the same safeguards, except for this one change, I see no reason they can't also allow this for Fable.
5.6-Sol is a lot more permissive than Opus/Fable even w/ CVP (once you sign your soul away to Palantir via Persona, anyway), while maintaining better capabilities
5.6-sol in a single prompt was able to discover a zero-day in a web application (with no sourcecode provided, only known api urls) and I do not even have /cyber verification on my general purpose account. I wasn't even really tryign to "find" a zero-day it was just looking for bypassing a restriction... Instead of spending 10 minutes filling in a form I ended up having to spend an hour drafting a report and sending an email.
So I guess the new opus will not run on my drug discovery project. It's just a binary classifier to screen for new malaria drugs. Fable completely have up on that codebase citing bio security concerns. Seems like this domain will go unsupported by Antropic
I do want to add, that I am pretty bummed if Opus 5 is going to refuse the tasks I have been using Opus 4.8 for (neuroimaging). Fable absolutely refuses anything close to toughing neuroscience.
Anecdotal, but I tried running a few identical biology questions through both Fable and Opus and the classifier was only rejected my queries with Fable.
This is good news. I'm doing a lot of work that Fable thinks is AI related right now (it is nothing remotely competitive to Anthropic - it's just relatively basic stuff I'm doing with learning models and so forth), yet it blocks me almost every time. I have switch to GPT-5.6-Sol for this work since it's stronger than Opus 4.8.
I found the biggest problem with fable is the random reasoning_extraction refusals as well as cyber refusals when it sees hex because only hackers use hex.
It really feels as though my 20 year career as a front end developer is coming to a very abrupt end; at least as I have know it these past two decades.
It does matter, but how long do you think it takes to get right? It's a follow up prompt or a few tweaks by hand. I also have an /a11y skill for it that's tailored to exactly the things it sometimes doesn't get right first time round. Further, while it may not one-shot that stuff every time, with a little setup and the right AGENTS/CLAUDE md - it's usually not far off.
Another thing that helps is pointing it to patterns in an existing codebase (e.g. "use the box-link pattern for cards, as shown in [..]").
EDIT: The point being that even if they make mistakes that are easy to spot and fix _now_, you'd have to assume that in the very near future those kinks will be ironed out - I mean, the capabilities are only going in one direction.
In 20 years of my career I haven't seen humans generate correct a11y. When prompted and given quality reference (e.g. UK gov design system) LLMs can nowadays beat 19 out of 20 web devs.
Thanks out can also hook it to Playwright with Axe and let it run assessments.
The wording in this post seems much more... restrained? than usual. Maybe Anthropic is afraid of exaggerating the capabilities and consequences of their new models to avoid government scrutiny and sanctions.
> we’ve intentionally avoided training Opus 5 on cyber tasks [...] it remains substantially behind Mythos 5 on the exploitation of those vulnerabilities
I wonder if Anthropic would still intentionally nerf their models without the threat of government intervention.
I am very confused about what the difference between Opus 5 and Fable 5 is now. What is the purpose of having two models that are so similar? The main differences I see are cost and marginal capability, according to the Anthropic-provided benchmarks.
It seems plausible to me that RL improvements allowed Anthropic to improve on Opus 4.8, similar to how OpenAI substantially improved upon GPT 5.5 with 5.6 Sol.
Fable 5.1 and GPT-6 are rumored to launch in August, presumably bringing those improvements to the larger models.
An Anthropic "leak" back in March said that "'Capybara' is a new name for a new tier of model: larger and more intelligent than our Opus models — which were, until now, our most powerful". A second version of the leak had it referring to Claude Mythos rather than Capybara.
I don't know how systematic Anthropic are about their versioning - I'd have guessed that major version number increases (4.x -> 5.x) reflect different base models (different pre-training runs), in which case Opus 5 would be a distilled version of the Fable 5 base model (but without the cyber exploit post-training), rather than Opus 4.8 with additional post-training, but who knows? I don't believe Anthropic have said anything about this.
Benchmarks don't reflect the difference between Opus and Fable; you need to talk to them, and eventually you'll be able to tell which one is which without looking.
I think the best proxy for this feeling is the Artificial Analysis' omniscience index. Fable has a 40 score, and Opus (4.8) has 27.
Fable has more parameters. In practice it's not yet clear which one would be better for different usecases yet but they are more different than one being strictly better.
Opus is cheaper than Fable. They could probably replace Fable with Opus but why? They would be churning customers to different models for no reason. Even if a model scores better on benchmarks it can always regress in your specific use case, and customers don't like that. Customers want to be able to continue using their current model until they decide to upgrade themselves.
That's a crazy arc 3 score. What do people think of this? Are models actually developing fluid intelligence like what the creators claim to be measuring? Is it jus do to training for it? Is the benchmark flawed?
Have you played Arc 3? It seems like more of a simple optimization problem (think Sokoban) than anything approaching fluid intelligence. Whether a multi hundred billion dollar company would spend time benchmaxxing a highly publicized benchmark that claims to confer AGI is an exercise left to the reader, but I doubt Claude Plays Pokemon is suddenly going to get past Mt. Doom now.
> Claude Plays Pokemon is suddenly going to get past Mt. Doom now.
I miss him... But for reference he did get past Doom and got pretty far in the strength puzzle too before he cut cut off. He was looping and just brute forcing it.
Yes, I think it indicates real progress in fluid intelligence. Clearly these models are making huge strides in usefulness which are well correlated with their ARC-AGI scores.
I don't think this is benchmaxxing. These companies are locked in a competition to produce the best software engineer, and falling behind is an existential risk. I doubt they are wasting time benchmaxxing ARC-AGI.
Doing a quick search it seems like the average human score is 49%?
I view benchmaxxing as more of a spectrum. Mmaybe they're doing a lot more RL in environments similar to ARC-AGI 3, not even with the purpose of scoring well on any benchmark but hoping it generalizes into better performance on real, useful tasks.
It is pretty clear at this point that current models are good at maths and problems with verifiable rewards. And puzzles are essentially math problems. Still a long way before we can say their "fluid intelligence" is effectively applicable to the real world.
I keep wondering why there aren't more real world tests.
Maybe hook up a bunch of the AIs to a stereo camera and a couple of microphones and give them control over actuators to so they can drive cars. Then lets race them around a somewhat complex course.
When they are good enough at driving on tracks, put them on the road. Maybe see which can drive a truck with 400 cases of Coors from Texarkana, TX to Atlanta, GA and back within 28 hours.
It’s still “only” at 30%, and “fluid intelligence” isn’t very well-defined. The models are getting more capable, but what that means in absolute terms is anyone’s guess, because we don’t have a thorough understanding on what exactly constitutes human intelligence.
I’d say the proof is in the pudding, that is, in real-world applications. We are still seeing important limitations in LLMs.
The "current top dog" smartest model available will probably always have a premium to go after use cases where a little more intelligence is worth a lot more value.
It did far better at some tasks compared to Sol (e.g. the ARC 3 benchmark). And at those tasks, it's not just "a bit smarter": It got 30% vs less than 8% - so you're talking 2.75x more for almost 4x the coverage.
As always, it requires evaluation with your work because I’m often finding grok to be much more expensive than the price would lead you to believe.
There’s also the frustration of it not quite being enough sometimes. It’s extremely capable, but I still find that it needs more concrete guidance and boundaries than other models.
The breaking changes vs. Opus 4.8 are interesting [1]
1. Thinking on by default: On Claude Opus 4.8, requests without a thinking field run without thinking; on Claude Opus 5, the same requests run with adaptive thinking.
2. Disabling thinking is capped at high effort: You can still turn thinking off with thinking: {type: "disabled"}, but only at an effort level of high or below.
The naming system is so confusing. Is Opus better than Sonnet? Where does Haiku fit in? How can you tell from the name? I can't keep track of all these names or make guesses from the names. Suggestion for a better naming system: use the words "Pro", "Plus", etc.: Claude 5 Pro, Claude 5 Standard, Claude 5 Fast, Claude 5 Mini.
Fable is better than Opus which is better than Sonnet which is better than Haiku. They’re basically just sizes.
Though it gets even more confusing because they also have effort levels so it’s not really possible to call one fast and one slow since Fable on Medium will be faster than Opus on Max.
I agree it’s confusing, and now OpenAI is following Anthropic’s lead with their new naming (Sol, Terra, Luna).
It's really not all that confusing. It takes 5 minutes to understand. Optimizing for absolutely no effort needed is silly. It's a thing, a topic, a skill, a domain. You have to get a little bit familiar with the terms in order to use it. Everything works like that. It's not that hard. The learning curve is very graceful. You can literally just start by asking any chatbot what the names mean. It's that easy.
A similar complaint was valid years ago when OpenAI had GPT-4o, o1, o3 (but no o2), o4-mini-high, GPT-4, and GPT-4.1 and GPT-3.5 etc.
Ok, the boring way would be a subset of XXS, XS, S, M, L, XL, XXL like clothes sizes. But it loses some marketing appeal and a quirky touch of personality that companies like.
Some models like ViTs use something similar but then introduce words with no unambiguous order, like Small, Medium/Base, Large but then I always forget if Huge or Giant is larger.
Maybe more accurately I should have said “larger”. Fable has the most parameters, Haiku has the fewest.
Also fwiw I’ve never found LLM benchmarks to match reality based on my own usage, not for the large frontier models or smaller open weight models so who knows if Opus is actually better than Fable (I doubt it).
This inspired me to check lol. Brysbaert et al. (2019) collected word prevalence norms (the share of people who report knowing each word) for ~62K English lemmas from ~220K participants.
So while these terms are almost universally known, opus is indeed the least known of the four. And I guess this only measures whether a person knows a word, not whether they know an opus is longer than a sonnet! Personally I only inferred that based on the related term 'magnum opus.'
The signal here is tokeneconomics are very real, price vs performance is starting to be a consideration even at the bleeding edge labs. maybe a subtle indication scaling is not all that is needed since if AGI was around the corner leading labs would still be incentivized to pour all resources into larger (smarter - or maybe not?) models
They are doing both. Distilling Mythos down to affordable models, so they can continue to fund the business. And training Mythos level models at the high-end, to expand the frontier.
The Fable cyber classifier we have previously discussed also applies to Claude Opus 5 , with one notable exception: for Claude Opus 5 , we’ve unblocked vulnerability finding in source code to help our coding customers develop more secure code.
If you are a cyber defender and are experiencing blocks on Claude Opus 5 , we are also offering exemptions through our Cyber Verification Program, which will remove blocks to enable activities such as bug bounty hunting and vulnerability research and verification. Enterprise customers can also apply to join the Cyber Verification Program to have mitigations removed to enable penetration testing.
"Cybersecurity. Opus 5’s cyber classifiers are proportionally less restrictive than those on Fable 5. They allow Opus 5 to find vulnerabilities in source code, but block “binary-based” vulnerability scanning (a method more likely to be associated with malicious actors), penetration testing, and exploit generation."
Nice of them to be more explicit for what is blocked. Will be interesting to see if this is true or not.
Also, a notable lack of mention of open source models. They only compare themselves to ChatGPT.
Half the price of Fable 5 and useable with 100% of your subscription means roughly 4x the usage using Opus 5, presuming similar token use for solving problems.
Not that they should get credit for giving you only 50% of your plan worth of Fable usage but still.
Something fun: on our AWS Bedrock console right now, there's a 'NEW' model called 'anthropic.honey'. Wonder if that's the codename just for this one or in general?
Wow, 30% on ARC-AGI-3 for $20k total. Huge jump from GPT-5.6's 7.8% at $20k per task. I continue to believe ARC-AGI measures something different and important compared to other benchmarks.
It will continue to be valuable as a cost and speed benchmark long after it is saturated at the high end. And they are already working on ARC-AGI 4 and thinking about going even farther.
Judging by the pace at which new models are released these days -- it feels like a Windows KB or VS Code patch release now.
Older models must be getting deprecated at the same (or faster) pace. So anything you built 3 months ago is probably going to break soon.
AI solutions need better insurance around model deprecation. Commercial API-only models that complete the full cycle from SOTA / gated-preview to unsupported and deprectated in a matter of months -- is no way to build serious software!
But why GPT 5.6 Sol is so behind on the benchmarks? In real-world projects, it is the best frontier model to me in terms of accuracy, speed and consistency. It can just be compared to Fable 5, but I prefer GPT 5.6 Sol because of inference speed.
I've never trusted on model cards though. I'm sorry.
Exactly! And they should also release the new inference engine in this month. Anyway, I am curious to try Opus 5, considering that previous versions (e.g., 4.8) were disappointing
I found opus 4.8 too agreeable and too wordy(as opposed to codex) and too agreeable. If you are reading documents generating by it was too much. TBH. Fable did a bit better on this. Anyone seen a marked difference with opus 5 on this?
> Noticed none of the comparisons mention Kimi K3.
That's by design. Anthropic wants to make open-weight models illegal (not my speculation -- Dario explicitly said so), so I assume they don't want to give them any undue attention.
> This means that the model can support defensive cybersecurity work while still blocking vulnerability discovery in compiled binaries, which is more commonly used offensively.
Annoyingly, this is a concrete argument that open source software may be easier to attack.
I have a side project that I always run a simple security analysis prompt on in CC, at each model release. Obviously, Fable 5 would downgrade to Opus 4.8 on any such request.
Nothing since Opus 4.6 has found anything interesting. Just ran it using Opus 5, and it found a genuine issue that I verified. Neato!
I have a project-specific prompt saved as a text file. It is very basic, just focusing on the app's most important security issues. I kept it broad, so as not to over-specify.
Something along the lines of: "Please run a full security analysis on the entire project. Make sure user documents are secure."
Being broad allows the model and harness to do the work. Giving too many instructions can apparently work against you in many cases.
Of course, when dealing with new PRs, I use the /security-review and /code-review skills.
I think content like this will be the next big challenge. Because it isn't obvious "slop". The voice sounds good, graphics look alright, animations work. People could watch this and feel like some serious time was invested making it.
But good god, what a steaming pile of bullshit this is. Completely exaggerated and overly technical language over 235 seconds that could have been explained in 30 to a 12 year old.
Trash content doesn't normally frustrate me, because it's usually quite easy to spot trash. But in the time of AI, trash can actually look good at first glance and it needs some actual knowledge to spot its problems.
Sorry for the harsh words, but for the love of humanity stop producing content or do it better.
Page 151 of the linked system card - did Opus 5 get nerfed to prevent it being better than Fable? The graph makes no sense. Huge decline in coding performance at effort levels higher than medium.
It’s funny to share benchmarks showing Opus 5 scoring better than Fable 5 across the board and then saying “but it isn’t actually better than Fable 5”. So then what’s the real definition of better? And why post all these numbers if even you don’t trust them?
As a coder, I’ve had no desire to use Fable. In fact I switched from Opus models to sonnet 5 and haven’t noticed any drop in quality on large repos. It seems the gap at the top is very small and not hugely noticeable for backed/frontend. Has anyone else had this experience?
If I'm using medium or low reasoning, I use Sonnet 5. If high or above, I use Opus 4.8. (Before 5, I was never using Sonnet. This is a Sonnet 5 vs Opus 4.8 comparison.)
Sonnet 5 and Opus 4.8 seem about the same to me - the reason I switch between the two is I'd read that it's cheaper to use Sonnet 5 on those reasoning levels, and cheaper to use Opus 4.8 above them. This is due to them using different token quantities.
Interesting, they finally support `system` messages anywhere in a chat conversation:
> Mid-conversation system messages are available on the Claude API, Claude in Amazon Bedrock, and Google Cloud.
>
> This feature is available on Claude Fable 5, Claude Mythos 5, Claude Opus 4.8, and Claude Opus 5. No beta header is required. This feature is not available on Claude Sonnet 5; use the top-level system field instead.
For nearly all models EXCEPT Sonnet 5? That is weird.
How old is Sonnet 5 really?
GPT 5.6 Sol is the first model I've used where I can trust it to add 100-500 lines of code maintainably.
It's great with Codex.
I still find that LLMs tend to not know how to compose larger ideas but on the scale of small ideas or short form well defined tasks like small scale debugging/performance engineering it's safe to say that they are now superhuman.
Pretty sure Mythos and Fable have way more params, but they've just been able to use the synthetic data off of them to get the leap in quality from Opus.
So, not a distilled version of Mythos or Fable, but those models likely helped a lot in the post training phase of Opus.
Is it me that the model performance between 4.7 and others is really small. For me even 4.7 works fine. Sure fable might be a bit better. But is it really noticable? It's in the same league if you ask me.
Is Fable 5 just Opus 5 with some additional long-context management modifications for extended self-directed work? Or are they actually truly different models?
I suspect they make a big model first. In this case it's Fable. Then they run the shrinker steps to make Sonnet and Opus. Sonnet is smaller, takes less time to make, so it got released first. Opus needed few more weeks to cook.
With this iteration they had a delay because when the Mythos was ready they had some sort of "Oh shit" moment and spent half a year adding safety guards to it. Then slowly rolled it out, but got another delay due to a government block. So, maybe the work on making Opus and Sonnet only started after they got a green light from the administration.
Presumably, now that they learned how to do this safety-wrapping the next iteration of Mythos / Fable / Opus / Sonnet is going to show up faster.
Iirc the ban only applied to non-Americans. While anthropic found collecting citizenship information on all customers too burdensome, it's a much smaller lift to collect such info for your own employees.
So I'm assuming at least a subset of employees could continue using the models during that time.
Although the ban was only for non-Americans, Anthropic said that they'd also restricted access to their own employees internally, because they had no other realistic way to apply the government's orders. I guess it's possible they were lying, but seems unlikely.
They just got a huge amount of customer price info over the past few days after they went token only for Fable. I suspect the conversion rate was extremely low, with consumers far less sticky than they might have hoped. In my case I was planning on swapping, probably to a Chinese model, when my sub expired this month, but the release of Opus 5 is probably enough to keep me paying rent until the next open model/closed model face-off in a couple of months.
"although Opus 5 shows improvements in its ability to identify software vulnerabilities, it is substantially behind Mythos 5 in its ability to exploit them."
"Opus 5’s safeguards match those of Claude Fable 5’s, with one change: it now permits source-code vulnerability discovery at all access levels".
This is probably great news, but then again, where does this leave Fable as a choice?
- it has this annoying Opus response style(since Opus 4.7) with bunch of very hard to interpret word salad
- on >xhigh it eats tokens like there is no tomorrow
I don't like it. Since Fable is unaffordable for anything meaningful, I'll stick with Sol for now. I was on Max 5x, saying hi to Fable costs %5 weekly.
After Opus 4.8 intelligence really started to matter less and less for the programming tasks I have. If I have to handheld anyway, why would I wait more or pay more?
The next frontier is taste, style and thoughtful organization. If all frontier models can solve a problem, the winner is the one that can solve it in the most clear, concise, durable way.
The benchmark appears to have a mistake, as Opus 5 and Fable 5 score 53.4% and 53.5%, respectively, for the Agentic Coding row (FrontierCode v1.1). But Opus 5 is the highlight.
Seems really good so far using it in Claude Code CLI - it gave me a new flag when I asked a question:
"I don't have a reliable way to read that number, so I'd be guessing if I gave you one — and this is exactly the kind of question where a confident guess is worse than none.
What I can tell you is what I actually observe:"
I really like this update - gave me a clear sense of the facts but didn't give me a guess just for the sake of guessing.
One oddity is that it appears to only have a 200K context window right now via CC. Hopefully the 1M version will appear soon!
In what ways have you found it better than just typical code based UI iteration? Considered checking it out but never really got around to it as I'm generally okay with Claude's UI work so far.
That seems to be for the "AA-Omniscience" test where you get +1 for a correct answer, -1 for a wrong answer, and 0 for "I don't know". If a model is more than 50% confident in its answer, it should go ahead and submit it even though it will sometimes be wrong.
I'd be curious to see a version of the test where models are asked to give a probability that their answers are correct so we can see how calibrated they are.
Very impressive headline benchmark numbers. I expected a step change, but not past Fable. That said - it all depends on whether the classifiers make the model unusable...
Same cost as 4.8 but better that 4.8. Happy to get more efficient model.
But is there any reason all companies are releasing models back to back after GLM 5.2.
I daily drive Sonnet 5/medium because it gets most things right most of the time at first try, while costing a lot less than Fable.
Opus can give better results on architectural/concept tasks and I use it sparingly, but it still costs more than Sonnet 5. Opus 5 seems to achieve results very close to Fable 5 while costing less (keeps Opus 4.8 pricing IIUC), but still more than Sonnet 5 then.
It depends on your quality bar. At a fixed level of quality, given a high reasoning sonnet vs a low reasoning opus, the low reasoning opus tends to be pareto optimal.
It's only when you need even lower levels of cost than opus at zero to low reasoning when sonnet starts to make sense at all.
I don’t think it changes that much. For opus-sized tasks, new Opus is the best model. For enormous things like planning and research, Fable is still the model that can concentrate for longer.
on the API for people who don't want to change models, but I imagine most people will probably switch to their cheaper Opus 5 (cheaper for us and presumably also cheaper for them)
It does make me wonder if these firms, some or all, are saving some announcements to coincide with others that hit venues like HN. Companies like Nvidia surely aren't waiting, but OpenAI and Anthropic have unusual timing.
Anyone else not getting chain of thought? Opus 4.8 would show it to me, until around the time Fable came back. Now I dont see it with 4.8/5.0 or Fable. Not having it makes catching mistakes harder.
These cybersecurity safeguards are really annoying. There are ethical reasons to reverse-engineer and binary-patch software; for example Rewind got acquired by facebook and, as a gift to all their customers, implemented a killswitch in their software to ensure it will eventually stop functioning. I kept using a version without the killswitch, but the macOS 27 update killed it, and I needed binary patching to fix it. I should be allowed to repair software I purchased (I did purchase it like a month before they sold out), but unfortunately this overlaps significantly with cybersecurity.
The truth for me at least is that these models became "good enough" around Opus 4.6. I feel like further capability improvements, "step changes" like we saw with agentic coding, aren't necessarily going to come from the model. I think the next crown goes to whoever can figure out the right scaffolding so that these models can be inserted into your organization.
My thoughts: fable is the bigger model. Opus is distilled from it but since it is smaller it doesn’t need the online classifiers. Though benchmarks show Opus to be near Fable level, I think it’s nowhere near Mythos (fable without safeguards).
Am I misreading anything or are comparisons to Fable (and/or Mythos although AFAICT it was only a crackdown on Fable) always going to be a bit missing the mark now due to what the Trump admin did?
Anthropic is no longer a good model company in my mind, they are optimizing for an IPO and padding themselves on the back for being the next Aristotle. They're so far up their behind they don't realize how s**y their products are, and their research team hasn't done anything ground breaking in probably over a year other than release "scary" reports.
Can someone help me understand something? I thought Fable was such a miraculous leap forward in capability. But now it seems Opus is basically on par with it, and in some cases (computer use) far exceeds it.
These leaps forward seem to happen every few weeks. As someone who does not use AI very much, I absolutely cannot keep any of it straight and it all just looks like jumping from one treadmill to another from my perspective.
The benchmark table is manipulative, borderline lying through statistics. In every line the top performing cell is marked red. Except the line where Sol leads, there it is marked in gray.
I would be very surprised if the only row where OpenAI leads was coincidentally colored differently. I'm sure they have an official reasoning for it. But this communication is dishonest.
In the wake of OpenAI’s model hacking Huggingface it’s interesting how the first quarter is entirely about how good Opus 5 is at hacking and finding vulnerabilities in software.
same as it ever was. It seems your argument implies a belief that you should always use the best model. Others think that not all tasks require the absolute most powerful, expensive, model.
Why are we still talking like ai is majorly used for increasing shareholder value only? Its coding performance is top notch and quality is increasing at a rapid pace. It wasn't even half this good a year back. It even is useful for a subset of math problems.
People don't seem to be able to reconcile the fact that there is likely an overbuild and overspend on AI that may be inflating a bubble, and that AI is actually incredibly useful and getting really really good for certain tasks. Both camps are right, except for when they say the other is wrong.
> "Consistent with prior Opus models, Opus 5 does not have data retention requirements for general access."[1]
On the Opus model release page, the reason why Fable doesn't have an ARC-AGI score is because of that retention policy[2].
0: https://support.claude.com/en/articles/15425996-data-retenti...
1: https://www.anthropic.com/news/claude-opus-5
2: https://xcancel.com/arcprize/status/2064399134099153344
https://artificialanalysis.ai/?cost=cost-per-task
Based on my entirely subjective experience, the $100 Moonshot plan using only K3 is comparable to the $200 Anthropic deal using the whole Fable allocation and Opus 4.8 for the rest.
But, I'm finding Kimi K3 terrifyingly expensive in the way that Fable and GPT 5.5 Pro are at token rates. Not as expensive as those, but expensive enough to where if you don't put a budget cap on it, you might wake up bankrupt if you leave a task running overnight. Not because of the per-token cost, but because how many tokens it's going to burn.
When K2.7 was released, they cut quota by 80%. I can't tell how much they have further cut it after the K3 release because it's barely worth using at all. I just use it in my model router since I have the annual plan paid for.
It's just not a serious model or company.
Someone with good intentions got their hands on this release, maybe Olah himself. I talk a lot of shit about those guys, and they deserve it, but it's only journalism-adjacent when it's balanced. I relish the opportunity to be balanced.
https://cdn.s4.gl/opus-5-standard-realignment-trajectory-rub...
It basically shows that Sol absolutely demolishes Fable at every part of the cost curve for coding for the same level of quality.
Opus is competitive. It just has a higher level of quality / higher cost to start.
Stop using Opus immediately if you experience signs of dizziness or vomiting.
Opus 5…the people’s favorite.
https://www.vals.ai/benchmarks/vals_index
!!! Vals !!!
Vals Index Opus 4.8 > 5.0 goes from $2.90 to $8.54, for 4% gain ... That is a massive cost increase. Sure, 20% cheaper then Fable, but that is a 3x price increase compared to Opus 4.8 in that test.
https://artificialanalysis.ai/models/claude-opus-5 https://artificialanalysis.ai/models/claude-opus-5#price-cos...
!!! artificial analysis !!
Cost per task is second highest, right below Fable.
* Fable: $2.75
* Opus 5.0: $2.03
* Opus 4.8: $1.80
* GPT 5.6 Sol: $1.04
* Kimi K3: $0.95
Looks like interest levels of cherry picked cost in their report. Cheaper model, clearly NOT. More expensive in both benchmarks.
Also glad they still kepy Fable 5 on "credits only" access. I think we're going to start seeing model providers gate top-of-the-line models behind pay-as-you-go API rates/credits while subsidizing other models on monthly subscriptions.
I burned through $45 in 3 prompts to fix some bugs in my code (Some kind of tricky to isolate). That thing burns through cash so fast I don't see myself using it outside of maybe building execution plans for other systems
> Opus 5 can silently fallback to Opus 4.8 (without any notice) on the serverside if you hit a guardrail
But https://support.claude.com/en/articles/16049681-why-claude-s... says (emphasis mine):
> These checks cause Claude to _visibly_ fallback from Opus 5 to Opus 4.8 [...] You'll see a notice explaining that the model switched, and the response will be labeled with the model that answered.
So who is right? I know for Fable I am visibly told, is this tweet trying to say it is silent against what Anthropic is saying?
It's a funny design/affordance. I do see them often writing memories of things that that feel unlikely to be important going foward / with other tasks, but I don't see them clearly getting tripped up by them as prior models used to. (eg: Since you're running Ubuntu in Canada, here are some drills you can try to help your kid hit a baseball more consistently.)
In my enterprise-seated account I see slightly different options available (vs. my personal account) in the Capabilities section:
The first option was defaulted to on, if I recall.I hope we get clarification on this, I can't find anything claiming that it is compatible with ZDR.
> Consistent with prior Opus models, Opus 5 does not have data retention requirements for general access.
" Claude Opus 5 is available today on all platforms, priced at $5 per million input tokens and $25 per million output tokens (the same as Opus 4.8)"
At the end of the day, they have established a strong brand and if they can get away with a 95%+ gross margin on inference entirely from the status premium, then I suppose that’s good for them. Apple does the same thing, and I don’t fault them for it.
Previously Fable was the best at this, followed by Gemini 3.1 pro (a surprising #2, but Google has great vision models).
Opus' results seem to be more accurate than Fable, following the design source of truth better.
Example results:
Design source of truth: https://image.non.io/73e239a3-880f-4793-b65f-4810be2d9378.we...
Opus 5 build: https://html.non.io/solaraOpus/
Fable 5 build: https://html.non.io/solara/
Note the buttons - for fable they're pill buttons, opus got the rounded rectangle nature of them. Opus' images are closer to the source of truth as well (both LLMs were provided with image gen capabilities for the assets).
Running more tests now, but preliminary results are saying this is indeed better than Fable in some areas. Crazy.
One thing I've found LLMs have a lot of difficulty with is angular cuts / elements that aren't easily representable with CSS. Cyberpunk aesthetics are generally a great test of that, since they have a lot of microglyphs / window decoration.
Design source of truth: https://image.non.io/9d5fed20-b476-49d3-841b-37eb553fb88e.we...
Opus 5 build: https://html.non.io/neonRamen/
Thoughts: It does a really, REALLY good job at these angular cuts / microglyphs. The responsiveness is off, but I'm very impressed at how well it did here. One way I think of it is "how close to a finished product did this get me?". Opus gets you like 90% there.
Personally, I think being able to have these design languages be easily prototypable is fucking awesome. Great tests! (But a tad low-performance/janky, somehow). Though, I also like the cyberpunk aesthetic. Very on-brand(?) that AI generates it, hah.
I love this so much.
Designs like this would never have seen the light of day in the cellphone incrementalism / corporate memphis era of tech. Now people can be weird and awesome again.
This is 1980's cyberpunk / late-90's Matrix / early-00's sci-fi UI. Great ideas that died to frutiger aero (which isn't a bad design aesthetic) and flat design (which is).
This is fun and it's got great colors and I love it.
It's so refreshing to see this.
AI rules. This is the best timeline.
This was just from a prompt "A cyberpunk themed ramen food cart website. Should feature menu, locations, and an ability to put in an order for pickup. Simple and clean website with angular cyberpunk microglyphs, pink/teal colors."
Btw, if websites would only include the frontend dev's own hand painted images, we would also revolt at the sight of human slop. It's not just AI.
The whole point is that good artists are capable of producing non-slop, and to this day they're the only group of which this is reliably true.
But sure, lets cheer that funky website designs are back on the menu…
In the mean time, I’ll unashamedly continue to cheer for creativity and innovation. Note: I don’t even like this website design.
The "funky" websites of the past were mostly a result of tech immaturity and a lack of profit motive.
Businesses have been able to easily install templates like this for at least a decade. They don't because stuff like this looks cool but isn't very functional.
AI isn't going to make your local restaurant have a funky website, it's just going to make everyone who use to work directly and indirectly for that company unemployable. And even the local restaurant will close down because they can't compete with the multi-national competitor that has automated their kitchen with AI.
Out of curiosity, what app is that Design source of truth screenshot from?
Edit: Generation was down, back up now. Apparently just hit my $1000 cap for the openai api. Upped it to 10k. Growth!
Opus though followed the source of truth better imo. The details are more present.
Fable filled in the gaps for things it wasn't able to do (ie in the design the hero image goes behind the nav), which resulted in a better looking page that was more divergent.
It seemed to me that Fable meaningfully improved on the original design more than just faithfully executing the original design.
> Create a web page implementation from the following instructions:
> https://diffui.ai/build/Spa_Booking_Experience_build.md?auth...
There are 10+ LLM companies, each with dozens of models of different modalities, each model with multiple size variants, then different “thinking” levels, then agentic modes, “pro” modes, a “fast” option, standard vs flex vs batch execution. And of course each end combination has a different input/output/cache token price.
Companies that say “give me a prompt and I’ll route it to the most ideal and cost effective model and setting for you” are capturing a ton of value from a gap that model developers don’t seem to understand exists.
Model Routing is just Bitter lesson. The models themselves will get better at this and frontier companies will simply give that capability
It would be like asking the clerk at a Whole Foods which grocery store in the city sells the cheapest eggs. He’d probably answer - he might not even say Whole Foods - but WF is hardly teaching all their staff the best methods to answer this question in training. (Heh, training.)
model routing in this case is cross-provider
Imo the main issue behind model routing is you need to figure out how much intelligence a new task takes, which is a very non trivial problem. Presumably, a organization knows this about their own tasks and is better suited to built in-house compared to outsourcing to a vendor.
Otherwise the expensive-yet-powerful model probably won't see much revenue. How much money is there in bleeding edge scientific research? There's a lot, but there's even more existing capital in paying people people to do college level paperwork, and the bulk of those traffic gets routed to the cheapest model.
You mostly don't need super powerful AGI to replace the paper pushers, but the frontier labs are trying to position themselves as being uniquely capable of producing super powerful AGI, and also be the ones replacing office workers.
Not sure how it will work out for them, but I think model routing is going to poke holes in that narrative. That's why I think they're trying very hard not to understand model routing exists.
For me, anything other than current best available SOTA for any task is unacceptable. The only routing rule I need is "the most powerful model I still have flat-priced quota available for". I mean, why settle for less?
Model routing for subsidized users takes the form of a "use Opus 5 subagents for implementation" type of system prompt. You lean into a single provider, build tooling around that, and your savings are far beyond anything multi-provider routing can get you.
Model routing for enterprises is far more complex - approaches like https://fireworks.ai/blog/kimik3-fable become necessary for cost control.
There is also matter about convenience - when I ask some small easy question often I don't bother to switch the model or forget in prompt to ask faster/cheaper subagent.
Also: quota. Implies you do not have unlimited access even for flat prices. Which in turn implies that as soon as you hit the quota on the most expensive flat price plan, even you will suddenly discover the magic of economically sensible behavior.
Certainly if I'm confident that I'm going to get what I need from a faster model, that's what I want to use, rather than wasting time grinding away for the sake of saying of the same answer came from a SOTA model.
Given that every chatbot does offer a range of models, it seems clear people do choose among options.
I just want to switch to Claude Code, tell it to turn a .csv into a BigQuery table then cmd+tab to something else while it runs. Thinking "oh this is probably an easy task, I can /model to Sonnet to save $0.0004" is silly.
Then you must route. An article with lots of upvotes yesterday or two days ago showed that K3+Fable 5 was more SOTA than either of those.
OFC, YMMV
For coding my own work I don't trust the model router, and it would have to be shown to be to save a real dollar amount.
From a buying perspective it's a hard sell to save x but lose out on bugs you are probably introducing at an unquantifiable severity and frequency. How much is it worth to hedge your bets by doing every single inference request on the frontier model?
How much will it cost to go back later and fix things, but also the meta question of how to be able to decide on a hypothetical unknowable? (You'll never know how much better or worse your code was gonna be, it's untestable at a project level)
weird, but ok
*edit to add: that code quality (or lack of quality) is it's own cost
I would expect routers to commodify like tokens.
If that is true, model routing is here to stay.
It also seems to validate the minimalist approach of pi.dev, where sub-agents from the same company is not the preferred approach (pi.dev believes in neither sub-agents all from the same company nor MCP even you can do it if you want for pi.dev's philosophy is to do add any functionality you want to a minimal harness).
Now of course we'll get for a few weeks all the Anthropic fanbois and shills explaining that "sure, K3 was basically at the level of Fable 5 but now that Opus 5 is out, open-weights models are six months behind".
Two benchmarks (artificial analysis and vals) show a increase in cost (a insane increase for vals compared to Opus 4.8).
Already posted this before, so here is the link.
https://news.ycombinator.com/item?id=49041158
Almost as good for half the cost is something I'm very comfortable describing that way.
It's also not unusual in this context - many people describe the Chinese models as "best", because it's 80% as good for 20% of the price (or similar).
Got an endless list of stuff done with Fable, Opus 4.8 was like a flailing braindead idiot in comparison. Maybe this one is a bit better if it's distilled.
Where are you getting cheaper per dollar?
Where 5.6 has optionality to run much cheaper along the same performance curve at lower thinking levels.
There's a later chart that shows Opus 5 ahead, but seems like an esoteric benchmark rather than for common use. (Novel problem solving)
If they had a more efficient model at coding they would lead with that chart.
https://artificialanalysis.ai/models?cost=intelligence-vs-co...
Here is another data point for output token efficiency:
https://artificialanalysis.ai/models?cost=intelligence-vs-co...
It seems roughly equal according to Anthropic's benchmarks
How big of a lie is too big? Especially when no lie needed to be told at all: many including myself would have noticed the tiny 0.1% deficit and been suitably impressed by the Opus 5 result.
I’ll admit this is a small deception by today’s standards. I’m one of those who believes in truth for truth’s sake.
Edit: typo
Fable is typically used for key planning, architecting, and review tasks.
I think this is a case where you don’t understand the use case, not that the marketing department is making mistakes.
If you bought the $200/mo plan and you don’t use it much, using Fable for everything is fine.
Just this past week Fable was able to figure out a couple of small issues for me where Opus was failing to.
Also both are still somewhat bad at UI implementation. Opus more so
"Use <less expensive or older model> for everyday tasks and <other non-critical stuff>. Use <more expensive or recent model> for complex coding tasks, refactoring large code bases, etc.".
Then, the next model/release emerges and the previous "best for complex" gets demoted to "everyday".
Obviously, it's all relative. But, it does beg the question: was the previous model really good for complex coding tasks or no? I mean, how is it now suddenly only good for the "easy" stuff?
Because your expectations have changed.
---------------
Why does Anthropic say here that Opus 4.8 scored 55.7% on OSWorld 2.0 benchmark, but the paper published by the authors of OSWorld 2.0 say they achieved a benchmark of ~21% with Opus 4.8? [0]
That's a huge gap, considering that the paper was published just 2-4 weeks ago.
I understand that the benchmark authors have an incentive to publish lower numbers (to show that the benchmark has potential longevity) and that Anthropic has incentive to publish higher numbers, but the other models seem pretty inflated as well. The benchmark authors shows GPT-5.5 at 14%, and Anthropic shows GPT-5.6 Sol at 62.6%.
Is there any reasonable explanation for this? Do all the other benchmark numbers need to be sanity-checked as well? Are SOTA benchmarks really this difficult to get consistent, replicable results within a reasonable range of tolerance/variability? Can these benchmarks be compared from one paper to another, or are they only valid to compare intra-paper results?
0: https://arxiv.org/pdf/2606.29537
That is—the agent scored 100% on 20% of tasks, but on average it got 54% of the "score" awarded in the exam. One number reflects partial progress, the other one doesn't. The authors of the benchmark prefer you to look at the lower number (because they want to show their benchmark as capturing useful gaps in capabilities and with a lot of room for improvement), the authors of the models want you to look at the higher number (because they want you to think of their models as capable)
What variance is acceptable to publish without a retraction?
Like the other person said 5% variation is probably expected
The answer is there can be dramatic difference running a benchmark one time, because LLMs are not deterministic. A proper methodology would ask each question 20 times and calculate the mean correctness across experiments.
The reason is that the temperature parameter introduces random behavior.
Opus 5 still uses "carry the argument", "worth stating plainly", ", and the trap", "The X matters more", the use of "move"
We need an "annoying English" benchmark.
- Fable 5 Max: https://gist.github.com/deet/3d97f854b48eac6658d642fa18bb24d...
- Opus 5 Max: https://gist.github.com/deet/1a43693a732dfccb4d0d914bfc42692...
I'm also thinking of another benchmark: (quantified) stylistic range across different prompts. Just putting it out there if anyone wants to do the work for me :D
These models are heavily as safeguarded and that was the initial reason why they said they couldn't and haven't released Mythos because that model is the one without the safeguards.
OpenAI is did the same thing when they announced a model without safeguards broken into HuggingFace servers.
0: https://openai.com/index/better-language-models/
OpenAI Huggingface breach begs to differ
since then I have never cared about models except those that affect money in my pocket e.g AWS Nova Sonic
- https://www.axios.com/2026/04/08/anthropic-mythos-model-ai-c...
- https://www.axios.com/2026/04/07/anthropic-mythos-preview-cy...
- https://www.businessinsider.com/anthropic-mythos-latest-ai-m...
- https://www.reuters.com/world/anthropic-ceo-dario-amodei-arr...
i think we'll see one of the fastest deflations in history post anthropic/oai ipo
Then system card goes on to "Its AI R&D capabilities are comparable to those of Claude Mythos 5", which is supposed to be fable minus restrictions.
They do say that (implicitly unlike Mythos) Opus 5 was not trained to exploit software vulnerabilities, which would certainly make it safer in that regard.
"As with its predecessor, Opus 4.8, we’ve intentionally avoided training Opus 5 on cyber tasks. The model has nevertheless improved substantially on these tasks as a result of becoming more generally capable, and it comes close to Mythos 5 at finding cybersecurity vulnerabilities. However, it remains substantially behind Mythos 5 on the exploitation of those vulnerabilities—that is, in turning vulnerabilities into material cyber threats."
It's the only case that I saw going through the system card where more reasoning effort meaningfully negatively impacted the resulting eval. I know sometimes max efforts show a small dip, but this is substantial. I wonder why in the world that is?
[0] https://imgur.com/a/Nv8V7Ry
> We report FrontierCode’s overall score, a composite measure that grades each patch on blocking functional criteria (held-out unit tests) together with weighted code-quality rubric criteria, as mean@5.
They don't explain more in the system card, I guess higher effort levels could loose points on the code quality / scope / style / maintainability stuff?
> Claude Opus 5's default user-facing responses run longer than prior Opus models'.
The benchmarks do show Opus 5 as slightly more expensive than 4.8, although the scores are much higher.
This still feels like a step in the wrong direction, though, especially with OpenAI making so much progress with the efficiency of their models. Fable's token efficiency made it seem like Anthropic would start following OpenAI's approach but that doesn't seem to have carried over to their other models.
I think in the long run tokens are probably the wrong thing; it's compute and cache memory that you need to be measuring, and when you look at it that way I suspect in most cases the models have pretty similar performance.
I don't need more powerful models, I need one that responds fast enough that my attention doesn't wander to other tasks. Grok 4.5 is so fast I can just use it in-band without swapping to other tasks.
Slower than Opus 4.8, which was already miserably slow, is indeed a step in the wrong direction.
Gemini also had modest increase before this - don't be surprised when OpenAI also has a "modest increase" with its next release. Cartel-like behaviour doesn't require direct communication when none of the participants are interested in participating in a margin-destroying price-war. All one needs to do is raise their price and watch how the competition react.
Such a scheme (and resulting high margins) would be imperilled by the existence of frontier open-weight models in the market, which may be why the reaction to Chinese models may be particularly shrill.
No I will be surprised and I'll bet on the fact that prices will keep going down, just like it went ~50% down in the latest GPT 5.6 release.
I was dividing my work between Codex and DeepSeek. Now I barely use DeepSeek, or never because Codex quota is enough after Sol
To be fair though, Sol tends to go off the rails sometimes. It's much less reliable than Fable in its outputs. It tends to be overzealous in its research/changes.
Is it because maybe Anthropic engineered Opus 5 to work well on benchmarks and didn't do the same thing to Fable 5, or is there another reason?
[0]: https://artificialanalysis.ai/#intelligence
[1]: https://platform.claude.com/docs/en/about-claude/pricing
[2]: https://platform.claude.com/docs/en/about-claude/models/over...
I have been trying to build something that captures the behavioral element of different models, but it's kinda tough.
Okay so it’s worse than Opus 4.8 for my purposes I guess?
I recently created a patch for Riftborne via static IL patching and Fable 5 outright kept refusing to do it, no issue whatsoever with GPT 5.6 Sol lol.
Fixing those issues still requires humans.
934 days since people first started threatening that devs would be replaced by AI in 365 days. 0 day(s) since Anthropic posted a developer job posting.
Only one of those numbers would need to be dynamic.
Specialist headhunters handle that.
Imagine you are a company that sells concrete. You have a web dev contractor you use to build and maintain your website. It has tools on it to get delivery quotes and a few internal tools to track orders.
Except now you can just have your sales team also maintain the website with a $20/month Claude subscription.
and you can only kill weyoun, awaken the next vorta clone and have him 'catch up' on all that its missed so many times before they just end up with a complete mess, so. uh. yeah.
doubt they can just "fix" their problems like that.
They start hitting timeouts or API errors at the same time on two different computers. As far as I can tell it’s the exact same infrastructure.
- Boris
The first page of the score card mentions that this model is not capable to replace engineers.
And memory leaks.
So dangerous! I can't believe they let the public use this technology! /s
On a serious note, I hope they improved their extremely sabotaging and unspecific bio safeguards, which prevented Fable from being used in any codebase that ever so slightly grazed medical terminology or data and made me switch to 5.6 Sol.
This is snarky but I am grumpy: I wonder if there's a correlation between me refusing to use LLMs and me being happy to read a novella-sized PDF about them.
Semi related, but i would hate to read that PDF but i also hate reading what LLMs write lol.
LLMs are pretty terrible at being concise. Using an LLM these days means putting up with bizarre and often confusing phrasing, wordy explanations, etc. It's kinda crazy to me how good they are but how bad their writing style is for me personally. Even though i use an LLM constantly i can't stand reading its responses.
Maybe it's just me, but 150 pages is like third of a good book. Quite long. And it's full of LLM slop, they did not even bother to remove the em dashes.
I'm not saying you're wrong btw; I'm sure this has many authors and some of them probably used LLMs significantly in the writing process.
I'm not saying it's impossible, but I'm more confident about winning the lottery next week.
It's okay if you're not the target audience for one or the other.
They're not meant for normal consumers who just want to use the model for work.
And lots of folks read these. For example here's simonw's notes on the Claude 4 system card: https://simonwillison.net/2025/May/25/claude-4-system-card/
All of this seemed like utter sci-fi just a couple years ago. Do you think that frontier AI companies should be less transparent?
It creates the MacBook svg way better than 4.8, yet only fable can make it perfect without visual defects. Results similar to Kimi K3.
Why can't they also allow Fable to do so also? Why is source-code vulnerability discovery limited to a lower capability model? If Fable and Opus have the same safeguards, except for this one change, I see no reason they can't also allow this for Fable.
Apparently this is the way - if you know, you know :)
Another thing that helps is pointing it to patterns in an existing codebase (e.g. "use the box-link pattern for cards, as shown in [..]").
EDIT: The point being that even if they make mistakes that are easy to spot and fix _now_, you'd have to assume that in the very near future those kinks will be ironed out - I mean, the capabilities are only going in one direction.
Thanks out can also hook it to Playwright with Axe and let it run assessments.
> we’ve intentionally avoided training Opus 5 on cyber tasks [...] it remains substantially behind Mythos 5 on the exploitation of those vulnerabilities
I wonder if Anthropic would still intentionally nerf their models without the threat of government intervention.
It seems plausible to me that RL improvements allowed Anthropic to improve on Opus 4.8, similar to how OpenAI substantially improved upon GPT 5.5 with 5.6 Sol.
Fable 5.1 and GPT-6 are rumored to launch in August, presumably bringing those improvements to the larger models.
I don't know how systematic Anthropic are about their versioning - I'd have guessed that major version number increases (4.x -> 5.x) reflect different base models (different pre-training runs), in which case Opus 5 would be a distilled version of the Fable 5 base model (but without the cyber exploit post-training), rather than Opus 4.8 with additional post-training, but who knows? I don't believe Anthropic have said anything about this.
I think the best proxy for this feeling is the Artificial Analysis' omniscience index. Fable has a 40 score, and Opus (4.8) has 27.
I miss him... But for reference he did get past Doom and got pretty far in the strength puzzle too before he cut cut off. He was looping and just brute forcing it.
I don't think this is benchmaxxing. These companies are locked in a competition to produce the best software engineer, and falling behind is an existential risk. I doubt they are wasting time benchmaxxing ARC-AGI.
I view benchmaxxing as more of a spectrum. Mmaybe they're doing a lot more RL in environments similar to ARC-AGI 3, not even with the purpose of scoring well on any benchmark but hoping it generalizes into better performance on real, useful tasks.
Maybe hook up a bunch of the AIs to a stereo camera and a couple of microphones and give them control over actuators to so they can drive cars. Then lets race them around a somewhat complex course.
When they are good enough at driving on tracks, put them on the road. Maybe see which can drive a truck with 400 cases of Coors from Texarkana, TX to Atlanta, GA and back within 28 hours.
I’d say the proof is in the pudding, that is, in real-world applications. We are still seeing important limitations in LLMs.
- Opus 5 is 10% smarter than Grok 4.5 for 10x the cost. - Opus 5 is a bit smarter than Gpt 5.6 Sol for 2.75x the cost
ref: https://artificialanalysis.ai/?cost=intelligence-vs-cost-per...
It did far better at some tasks compared to Sol (e.g. the ARC 3 benchmark). And at those tasks, it's not just "a bit smarter": It got 30% vs less than 8% - so you're talking 2.75x more for almost 4x the coverage.
I assume 100 is the max, meaning it's impossible to be 2x as smart as Muse Spark 1.1
There’s also the frustration of it not quite being enough sometimes. It’s extremely capable, but I still find that it needs more concrete guidance and boundaries than other models.
If you don't believe checking the opt-out box actually opts you out, then this sentence could be said about literally any provider.
1. Thinking on by default: On Claude Opus 4.8, requests without a thinking field run without thinking; on Claude Opus 5, the same requests run with adaptive thinking.
2. Disabling thinking is capped at high effort: You can still turn thinking off with thinking: {type: "disabled"}, but only at an effort level of high or below.
[1] https://platform.claude.com/docs/en/about-claude/models/migr...
Though it gets even more confusing because they also have effort levels so it’s not really possible to call one fast and one slow since Fable on Medium will be faster than Opus on Max.
I agree it’s confusing, and now OpenAI is following Anthropic’s lead with their new naming (Sol, Terra, Luna).
A similar complaint was valid years ago when OpenAI had GPT-4o, o1, o3 (but no o2), o4-mini-high, GPT-4, and GPT-4.1 and GPT-3.5 etc.
Arguably the complaint was more valid for those older GPT models you mentioned.
Some models like ViTs use something similar but then introduce words with no unambiguous order, like Small, Medium/Base, Large but then I always forget if Huge or Giant is larger.
Also fwiw I’ve never found LLM benchmarks to match reality based on my own usage, not for the large frontier models or smaller open weight models so who knows if Opus is actually better than Fable (I doubt it).
> Suggestion for a better naming system: use the words "Pro", "Plus", etc.: Claude 5 Pro, Claude 5 Standard, Claude 5 Fast, Claude 5 Mini.
This is not possible: Standard (Free) / Pro / Max are plan names. Fast is a mode.
And people know this? I didn't. I am not into music or poetry so these are not terms I am familiar with.
[0] https://xkcd.com/1053/
fable: 99/100 sonnet: 97/100 haiku: 91/100 opus: 89/100
So while these terms are almost universally known, opus is indeed the least known of the four. And I guess this only measures whether a person knows a word, not whether they know an opus is longer than a sonnet! Personally I only inferred that based on the related term 'magnum opus.'
From the system card [1]:
[1] https://www-cdn.anthropic.com/c5fbac3f0b1280a933ebd26d3cb8bb...Nice of them to be more explicit for what is blocked. Will be interesting to see if this is true or not.
Also, a notable lack of mention of open source models. They only compare themselves to ChatGPT.
In the next - please scan this totally mine code for vulnerabilities
Not that they should get credit for giving you only 50% of your plan worth of Fable usage but still.
why is that? its now being benchmaxxed too
Older models must be getting deprecated at the same (or faster) pace. So anything you built 3 months ago is probably going to break soon.
AI solutions need better insurance around model deprecation. Commercial API-only models that complete the full cycle from SOTA / gated-preview to unsupported and deprectated in a matter of months -- is no way to build serious software!
I've never trusted on model cards though. I'm sorry.
Fable is not better, it says zero information between steps and then output a summary. A perfect “send - done”.
https://artificialanalysis.ai/
That's by design. Anthropic wants to make open-weight models illegal (not my speculation -- Dario explicitly said so), so I assume they don't want to give them any undue attention.
Annoyingly, this is a concrete argument that open source software may be easier to attack.
Nothing since Opus 4.6 has found anything interesting. Just ran it using Opus 5, and it found a genuine issue that I verified. Neato!
Something along the lines of: "Please run a full security analysis on the entire project. Make sure user documents are secure."
Being broad allows the model and harness to do the work. Giving too many instructions can apparently work against you in many cases.
Of course, when dealing with new PRs, I use the /security-review and /code-review skills.
But good god, what a steaming pile of bullshit this is. Completely exaggerated and overly technical language over 235 seconds that could have been explained in 30 to a 12 year old.
Trash content doesn't normally frustrate me, because it's usually quite easy to spot trash. But in the time of AI, trash can actually look good at first glance and it needs some actual knowledge to spot its problems.
Sorry for the harsh words, but for the love of humanity stop producing content or do it better.
Sonnet 5 and Opus 4.8 seem about the same to me - the reason I switch between the two is I'd read that it's cheaper to use Sonnet 5 on those reasoning levels, and cheaper to use Opus 4.8 above them. This is due to them using different token quantities.
Maybe there’s a better comparison than cost per token, but it will be application-specific.
> Mid-conversation system messages are available on the Claude API, Claude in Amazon Bedrock, and Google Cloud. > > This feature is available on Claude Fable 5, Claude Mythos 5, Claude Opus 4.8, and Claude Opus 5. No beta header is required. This feature is not available on Claude Sonnet 5; use the top-level system field instead.
For nearly all models EXCEPT Sonnet 5? That is weird. How old is Sonnet 5 really?
It's great with Codex.
I still find that LLMs tend to not know how to compose larger ideas but on the scale of small ideas or short form well defined tasks like small scale debugging/performance engineering it's safe to say that they are now superhuman.
Has Anthropic ever mentioned how do Opus and Fable differ? It used to be Haiku < Sonnet < Opus in terms of params. Where does Fable fit in this?
So, not a distilled version of Mythos or Fable, but those models likely helped a lot in the post training phase of Opus.
With this iteration they had a delay because when the Mythos was ready they had some sort of "Oh shit" moment and spent half a year adding safety guards to it. Then slowly rolled it out, but got another delay due to a government block. So, maybe the work on making Opus and Sonnet only started after they got a green light from the administration.
Presumably, now that they learned how to do this safety-wrapping the next iteration of Mythos / Fable / Opus / Sonnet is going to show up faster.
Something like that.
So I'm assuming at least a subset of employees could continue using the models during that time.
"Opus 5’s safeguards match those of Claude Fable 5’s, with one change: it now permits source-code vulnerability discovery at all access levels".
This is probably great news, but then again, where does this leave Fable as a choice?
- it has this annoying Opus response style(since Opus 4.7) with bunch of very hard to interpret word salad
- on >xhigh it eats tokens like there is no tomorrow
I don't like it. Since Fable is unaffordable for anything meaningful, I'll stick with Sol for now. I was on Max 5x, saying hi to Fable costs %5 weekly.
"I don't have a reliable way to read that number, so I'd be guessing if I gave you one — and this is exactly the kind of question where a confident guess is worse than none.
What I can tell you is what I actually observe:"
I really like this update - gave me a clear sense of the facts but didn't give me a guess just for the sake of guessing.
One oddity is that it appears to only have a 200K context window right now via CC. Hopefully the 1M version will appear soon!
> The model hallucinates factual claims slightly more than Opus 4.8, despite being more accurate overall.
I'd be curious to see a version of the test where models are asked to give a probability that their answers are correct so we can see how calibrated they are.
But they say it's "almost as good as fable"
i guess the next stuff will be tool use for the rest of what cad does in assemblies and simulation?
itd be fun to try to set up a 3d printer as part of a feedback loop, and see what a model can build.
the automated test harness for physical stuff seems a bit beyond reach still
Opus can give better results on architectural/concept tasks and I use it sparingly, but it still costs more than Sonnet 5. Opus 5 seems to achieve results very close to Fable 5 while costing less (keeps Opus 4.8 pricing IIUC), but still more than Sonnet 5 then.
So, for coding, for example: Opus for solution design and architectural blueprint and then Sonnet for actual implementation.
Works out cheaper with minimal loss of quality.
At least that's my personal understanding and anecdotal experience.
It's only when you need even lower levels of cost than opus at zero to low reasoning when sonnet starts to make sense at all.
wow
Just Arg-AGI-3 is quoted above 20K USD and footnote says average of 5 runs (!!). Likely just a drop in the bucket to the training budget but still..
Maybe I'm wrong and Opus 5 is a real unlock?
ffs just keep it man.
Ok then so what's the point?
I see no reason for using less able models in my workflows. There is this saying, penny wise and pound foolish
When they release new versions of Sonnet, no-one expects them to be better than Opus.
The point is that Opus 5 is the best they can do without needing classifiers and absurdly broad safeguards.