There were more than a few "leaks" saying it would be called Fable 5.5, so I'm curious if a) the leaks were just engagementbait or b) Anthropic changed it, despite the fact that Fable 5.2 being better than Opus 5.5 is confusing.
Surprisingly the model is available immediately in my Claude Code: usually there's a delay, either for rollout or safety.
Somewhat plausible values: 5.5 (to harmonize with Opus), 5.3 (we're skipping a minor version because...), 6 (because we're at the top of benchmarks again).
I agree that's a lot of money for many people. But I just can't wrap my head around complaints about $200 a month to access that must computing power. It's worth $12,000 of API tokens (and that's without accounting for the three resets they granted today to many subscribers).
I would suggest calibrating what "a lot" means. In terms of dollars per unit of intelligence, it's extremely cost effective.
If Anthropic were telling you you were getting 500,000$ worth of compute for 5,000$ a month, would you still consider this a bargain?
People don’t care what the costs for the LLM providers are. They are paying for a service. 30$ a month is reasonable for a lot of devs. 100$ is a lot for many. 200$ is unreasonable for personal usage for most.
Everybody understands the economics are all skewed and this tech requires a lot of resources / compute. There are still absolutes in terms of pricing that are not acceptable/affordable by most.
Maybe I'm too cynical, but I've never trusted the worth $x of API tokens. It's like seeing an advertisement say $299 of value now for $19.99. It's not an actual number right? that $12,000 of "API Value" is just some made up marketing number isn't it?
I'm open to learning more, not trying to ask a rhetorical question.
What you need for real engineering and scaling is cheap and just enough smarts. I've found GPT 6 Luna to be cheap enough that I can pretend the cost doesn't exist and it's smart enough that I can get real work done. This let's me scale it in ways that I can't scale frontier models. And funny enough, the LLM I have running on my 5070 Ti is just about comparable to what I get from Luna. Unless frontier intelligence gets waaaaaay cheaper I basically don't care about them.
> the number of Fable resets you get depends on how long you’ve been a paid subscriber [...] Subscription length is measured from the date of your first paid bill. If you canceled and came back, the time since your first paid bill still counts.
> These resets expire on March 8, 2027 at 11:59 pm PT.
tl;dr if you joined when Opus 4.5 became a thing, you got 2 resets.
Maybe it's just me, but I found Claude's writing style terrible compared to GPT or Gemini. It's "smart", but the way it communicates seems somehow off compared to other models.
Did you find that to be true for Opus 5.5 as well?
I couldn't use prior Claude models without rolling my eyes at the writing every few sentences, Astra was refreshing and then Opus 5.5 made even Astra look terrible in comparison.
Opus 5.5 got better for me but it still has some problematic behavior.
Interestingly, you can have informative conversations about the LLM linguistic tics with the LLMs themselves. One word that chatgpt used to describe what I was observing was linguistic anaphora (not rhetorical). It’s basically shortening/compression achieved via pronoun, and Claude does a lot of it. Worth reading about. A lot of it seems to do with how the new models are trained for brevity.
Another pattern I’m starting to become very sensitive to is Claude making up on the fly jargon for something under consideration, especially in code review. This morning for example it referred to “a claim” out of nowhere, and when I questioned it, it admitted that it was using the noun as shorthand for a record getting dequeued (“claimed”) inside of an sql transaction. It makes sense if you already know what it is referring to but in a code review it’s totally incomprehensible on first read, and ultimately counterproductive.
I really didn't like Opus 5 and was vocal about it. Opus 5.5 is noticeably better, but literally just yesterday I was fighting parsing some summary it had given me and it was a bit like the old days. IOW, it is better but still has some room to improve.
"You can tell an AI model to write in a different style" and "the AI model will actually write in the style you tell it to" are two different things that often don't intersect
In my experience, no amount of output style customization gets it anywhere near the output I'd like.
I've had more success running a prose linter[1] with a ruleset[2] that gets rid of most of the egregious stuff. You can also pretty easily write custom rules for it.
And sometimes, when I'm fed up enough, I just throw all the prose in the codebase at Antigravity/Gemini to rephrase. The result is so much more readable.
I had the same feeling. It's like every sentence becomes a struggle to parse: referencing things that haven't been introduced, using too many pronouns, the lack of verbs and the passive tense... These past few days it's felt more like Opus 5.1 than 5.5.
I just had Claude Code with "Opus 5.5" ignore one of my standing instructions for the first time and then proceed to generate an Opus 5 style verbose and barely readable commit message. I questioned it about Fable and Mythos and it has no idea what those were, yet they should have been in the Opus 5.5 knowledge cutoff.
On the web with both high and xhigh, it had knowledge of both.
The best programmers I've met or read throughout my life are phenomenal communicators. I'd even go so far as to say it's likely one of the reasons they're so good at programming. After all, the majority of our job is converting human requirements in lossy language to fleshed out behaviors in formal, lossless language.
Heck, some of the most common debates are on the semantic clarity of abstractions or syntax.
I've found Claude's writing between 4.6 until 5.0 to be unacceptable if you are someone who is juggling 4-5 Agents. The Sol 5.X models all seemed to have great writing. Now with Opus 5.5 and Sol 6.1 I've found the trend is reversed -- Sol is now unbearable and Opus is workable again. Whatever caused the issues for Anthropic seem to be hitting OAI now.
sol is ok. it still such a superthinky model that it gets tangled up trying to present way too much? i struggle to connect with it often, have ti ask it to try again.
terra has been prettt good for me for me, when i need something looked at an explained to me.
i dont ask for much non-technical output though, which i expect yeah is its own cans of worms.
My own experience having used 6.1-Sol/6-Astra and Opus 5.5 extensively, Opus is more verbose and chatty in it's user communication (not code though, code is usually more straightforward). Multiple times I've gone "I'm not reading all that shit" when looking at Opus 5.5 output. This was also with the concise mode enabled in Claude Code.
No its just straight up terrible. I tried out Gemini Argon when it released and my god did it write beautifully. Even the 3.8 Flash models are way better at prose, i do a lot of automatic scriptwriting for youtube and 3.8 flash is leagues ahead of anything from anthropic/openai
I enjoy a lot Argon, maybe more than chatgpt, but damn, like all Gemini models, lots of hallucination and sycophant to the max, so a bit useless for agentic coding.
Yep I find the way Opus 5.5 talks to me really annoying and irritating. Quite condescending and sanctimonious.
I prefer Gemini for day to day use as I find it a lot more agreeable to be working with all day every day, and Gemini is pretty much on a par for coding Vs Opus 5.5, at least for my needs.
I always thought it's the other way around. GPT-6-Astra is very very defensive and abusively spam clarifications to whichever claim it says, most of the time not important or relevant at all.
Almost every response it’s telling me “I could not verify X” where X is something that doesn’t need verifying. I was doing some game work yesterday filtering through some historical data, and it tried to put caveats about data accuracy in the game’s UI! If Fable made Age of Empires you wouldn’t be able to play it because the entire screen would be full of footnotes and caveats.
Very interesting that they set the flagship's version number to 5.2, but their penultimate model is 5.5. Do they want people to to continue to think that Opus is based on more advanced technology? The decision these AI companies make continues to baffle me.
FrontierCode 2.0 is such a poor benchmark for Gemini since it's the only model forced to use Devin CLI while other models get to use their preferred harness.
I wonder why that team singles Gemini out like that.
*Model 3 Early, not Model 3. And Model 4 was made available to employees ~6 weeks ago, so at the current pace it's very likely they're using something even better now.
I have never been thinking harder than I do these days: Engaging with mountains of agent-generated text, weighing decisions and trade-offs, correcting the agent's half-baked ideas and misunderstandings, reviewing the bizarre code that they generate... If I'm not just sick (and I'm pretty sure that I am), then what I have is probably closer burnout than atrophy.
Tbh, its both. Working with multiple agents is exhausting and definitely burns you out but you also do lean less on your ability to push through boring or difficult problems. I’m finding that with lighter AI involvement I am more invested and less burnt out.
Nerfing an existing model is the biggest crime in this space. I can’t imagine Dario okaying this because it’s what will immediately destroy your brand.
“Testers have found that Fable 5.2 and Mythos 5.2 communicate more clearly than past Fable and Mythos models. They generate outputs up to 30% faster than Claude Fable 5.1 and Claude Mythos 5.1, and they’re available at a lower price: $8 per million input tokens and $40 per million output tokens.”
Honestly incredible that they’ve managed to beat all benchmarks, while making it cheaper. Always happy to see improvements offered at lower costs.
Will be interesting to see what OpenAI does in response
I'd first check per task costs. I don't know, maybe it's just a fluke, but on my few tasks I ate 2 resets in under 3 hours, on the pro 200 plan. It goes quickly. (Granted, it aced the tasks)
Doubling Remote Labor Index (RLI) - at 41.3% compared to around 20% for the other models - I wonder what kinds of new agentic work this is going to unlock.
Those theoretical physics papers look like the real thing (i.e. not mathematical physics like the OpenAI Yang-Mills result). It looks like AI is finally branching out of math.
Interesting that there are no authors listed other than Claude. Also interesting that they published on "Hexagon" instead of Arxiv. I haven't heard of it before. Is it an AI-friendly alternative to Arxiv or something?
With early handouts of Qwen4 starting to be reported 5 days ago ... so if they're good, they would have had some hours to evaluate Qwen4 against Claude 5.5 and, then or right after was when Anthropic decided to put out 5.2 ...
So yeah ... benchmaxxing and all that. BUT nevertheless, the published results seem to say that Qwen4-Max appears to match Opus 5.5 on some benchmarks, and Qwen4-27B (probably by going even harder than Qwen3.8-27B on thinking/inference time compute) can beat Sonnet 5.5 on some things.
I guess they're not taking any chances. Keep in mind that Anthropic has declared they have non-public models (it's in their IPO filing). These are referred to by the shorthand names Model 2, Model 3, and Model 4. Essentially 4 is better than 3 is better than 2.
Fable 5.2 is "2.5" so to speak. So Anthropic was holding back, but not anymore. And now they can dial the performance up once more with what they have ready, and probably once more after that.
Its incredibly easy to roll out a compute instance to X% of users, or route traffic between compute instances
so not everyone will have the same experience. If I was saving on costs that would be the easiest way to do it without deploying, without any nefarious intent. The persons in charge of that likely don't even know about how individual users are impacted
At what point does this become false advertising, or even fraud? Imagine paying AWS for a high-performance server, but 5% of customers who order one quietly get a slower, cheaper server and still pay the full price. That's a serious claim and makes me skeptical Anthropic is doing something like that without more evidence.
I think they could easily reduce several parameters to slightly reduce cost (something like "it's the same model, a few less passes, but quality is still 95%+").
Or switch to a faster, cheaper cache mechanism or something that's almost as good, but much cheaper to run, in the name of improving efficiency and keeping up with demand. This switch to cheaper mechanisms is what it feels like to me.
It's a problem for us. When they nerf the model our productivity really drops (due to the specific use case and process we have). We've started using OpenAI and will gradually use it more and more if this continues.
"Nerf tracking" is adversarial and exploratory. It's not something Anthropic wants anyone to do, and it's not something that the "nerf tracking" designers are necessarily doing in a foolproof way. Don't assume too much from them either way.
Sometimes you just have insufficient evidence to be sure of something either way, and nobody gets to win the argument or shut somebody else down.
In this case, some have reported analytical work[0] that goes beyond "gut feel" anecodotal evidence, but even that's not going to conclusive and may not be trustworthy at all.
These nerf trackers do not work. I wish people would stop posting them.
They don't detect a quantized model until the quantization gets down to 4 bits or less.
Humans absolutely do detect quantization well before 4 bits...
Try it yourself if you want. Try a small local model, both quantized and not, and I would be very surprised if it isn't immediately obvious which is which.
But the claim here is "the models are nerfed before a release", not "they A/B test quants / efforts". If it were A/B testing, it doesn't really make sense to correlate it to release windows at all.
They might want to see how people react to a new model depending on whether they've been nerfed.
If the nerfing is lower effort (seems to be, presumably why they hide the CoT) then they may be wondering if they can get away permanently with lowering effort on the new release provided it gives a wee improvement.
For sure, myself and colleague saw this nerfing. He said - new model coming. It was incredibly obvious. I also started getting a lot of "How's Claude doing today" messages at the same time, which is presumably to get feedback on the A/B
I strongly believe it will be shit like this that will have people abandon them. They're being way too cocky for a company that pretty much didn't exist 2 years ago
If this turns out to be accurate, I think that would be a pretty unethical thing for Anthropic. Even if it is just a "we had to allocate compute to this new model release".
It should be fraud to swap out the model. We need a term for this, something like a bait-and-switch.
I believe fraud is a sufficient term. You claim to be selling X but you actually sold them Y. Bait and switch is a selling tactic to avoid fraud where they make you believe product is X and then when you go to buy they tell you X isnt available but they have Y. While also unethical, it is not fraud because they did tell you what it is you are buying before you buy it.
if they just straight up dont sell you something you bought then its fraud.
I think there is a ripe opportunity for expanding the definition of fraud to include changing the "ingredients" of a named product on the market. It is not limited here, we see it with food, and even in the medical field
(I will die on the hill that medical fraud was committed when mRNA therapy was used as a treatment for covid and it was 'sold' to the public as a "vaccine" despite not being the same thing as any vaccine before it. The definition of vaccine was expanded to include it without sifficient public disclosure to warrant informed consent imo. This is not a statement about conspiracy or the efficacy of vaccines or a moral judgement on whether it was for the greater good or not.. it was simply an act of fraud that robbed the public of proper informed consent by making people believe it was X when it was really Y.)
So yes, I think there is a large need for the public to at the very least put a strong handle on the ability to change what a product name means. It is a trivial problem to solve considering it is only a problem due fraud or fraud adjacent activities.
heres the solution: You have your product name, and then any changes to it REQUIRE a new version name. Like how you have Coke and Diet Coke, only formalized such that Coke cannot legally replace the ingredients of regular Coke with the ingredients of Diet coke and still call it Coke. With Software they could at least use version numbers and release modifiers. most of the time its built in already, but it needs to include the full stack. if they change anything that has an interaction with the user it should change the label seen by the user for the product.
You don't buy a hard quantity of an LLM you buy access in general.
A Opus 5.5 is not a Opus 5.5. Its Opus 5.5 of some checkpoint.
mRNA btw. was well known how it worked, it wasn't hidden and some people waited for the non mRNA version because of it. I never seen or read the argument you make and I do not get it.
A vaccine is something which protects you from an infection. What definition do you think changed in humans? That we make the protein? How is that so different from viral vector vaccines?
The consent is a certain level of trust from the society to a group of researchers limited by being humans and having to act in a certain time frame.
Nontheless, I do agree I would like to have a fingerprint and proper changelogs etc. for products I buy including digital and analog. But this has a downside as usual: The smaller you are, the more it costs you to (probably) register, maintain and change.
You know the good thing about Claude etc. right now? It doesn't matter. My invest is not 4 digits and there are that many changes, i'm quite happy to have been upgraded for 'free' from Opus 4.8 to Opus 5 (what do you call this upgrade, which was significant better?)
I'm pretty convinced that those who have constantly been posting these complaints, probably for every single model release out there, are indeed wrong.
> And not just in this instance, nerfing has been complained about in a cyclical fashion for many months.
No, it gets complained about constantly even if there is no model coming out. A completely unrelated thread on Claude will feature comments like "can I just say that [model] has suddenly become useless this week?" Merely asking that question is bound to get people agreeing with you because we are evaluating all of this on vibes and placebos.
I too feel that way sometimes, I don't think it's ridiculous that Anthropic might silently downgrade you as demand warrants. But asking others if they "feel" something I feel is less than useless.
What people claim is not "some experiments degrade performance", it's "Anthropic nerfed the models [because they are evil and do not like the user etc]".
Like most experiments do not save the company money, there was an experiment to test the new fable, people got to use it at opus-5.5 usage. But this experiment clears the thinking blocks when people go back to the vanilla model; not sure if Anthropic overlooked this, but idk, it does not sound like "nerf on purpose".
There are experiments to adjust the effort level, at the moment the ones that I've seen switch to a higher effort. Again this is not a nerf; might be annoying perhaps, the model does start to behave differently (and consume more tokens). But it's not a nerf.
Then there is apparently one experiment that replaces opus-5.5 system prompt with the one from opus-4.8; the model is the same; this experiment is perhaps a bit more questionable, but still it's not "anthropic nerfed the models".
If people think they are getting nerfed they should check what experiments are enabled (claude.json and session logs).
> "We need a term for this, something like a bait-and-switch."
The media coined the term "dual quality" for a European food scandal a few years back, in which conglomerates (including Coca-Cola) adulterated products with cheap, substitute ingredients, if they were destined for distribution in specific markets (mainly, lower-income).
The packaging exteriors were the same; and there were (I think)? broadly no price differences. It was essentially what the software industry calls "A/B testing", except with sneaking things into what people eat.
I don't have any good reading links on hand; here's one source,
idk, it hasn't felt nerfed at all to me, and it's not like i'm just doing trivial CRUD webapps either. opus 5.5 is very successfully porting my phone to postmarketos. claude just got the touchscreen up and working (and made a little program for me to test it, and it's legit).
I at least didn't see this. I had a baseline evaluation from before Nov 3 on Opus 5.5 medium, so I re-ran it this morning.
Eval context: 30 bug fixes and features evaluated by tests the agent didn't see. Ran on Opus 5.5 at medium effort in Claude Code.
20 of the tasks were last run on Nov 1, twice each. Nov 1 solved 36/40, and today solved 19/20. Thinking tokens per task were 0.95x Nov 1’s and output tokens 0.97x. The one significant change was speed: median time to first token decreased from 3.0s to 1.7s.
Caveats: one run per task today, medium effort only.
Personally, I've found myself getting lazier recently, trying to get a single session to handle tasks that should be broken into multiple sessions, etc. I've suspected claims of nerfing might stem from changes in user behavior. I'm glad more disciplined checks are being run.
This is an example of LLM text that doesn’t make sense.
Paragraph 1 says there were 30 bug fixes.
Paragraph 2 first says they tested it on 20 of the “tasks”. But why is the denominator 40? Where did that come from? It should be 30.
It goes on to mention another test, this time with 20 tasks. Again, it’s unclear where this is even coming from. Based on the first paragraph it should be 30 not 20.
I think it's confusing and incomplete, but not wrong. Paragraph 2 says the Nov 1 test included 20 of the tasks and ran each twice. That's the 40 denominator.
Why paragraph 1 even mentions the full set of 30 tasks if 10 of them are irrelevant to the comparison, I don't know.
I'm not sure if it's LLM text, or somebody who has been reading way too much LLM text and is now corrupted into mimicking it. My guess would be a mix of human and LLM. "...evaluated by tests the agent didn't see" sounds very Claude. The first line (2 sentences) sounds human to me. Everything after sounds AI, though the last line is ambiguous.
It's kind of sad how our meat brains are getting non-consensually trained on this categorization task....
It stands as written? The previous test had 20 tasks run twice each, so denominator 40. The recent retest did the same 20 tasks but only once each, as mentioned in the caveat. The 30 is presumably either a typo or a separate count (like the test had 20 tasks related to code with 30 bugs/features).
Also - LLM text? If GP is confusing it's because it's overly succinct and might have typos, which are not exactly the hallmarks of LLM text
Benchmarks don't do a good job at capturing the difference in quality of frontier models over smaller models.
Opus 5.5 was so good that if it wasn't nerfed - at least just looking at the benchmarks - the difference in cost here does not seem to justify using this for anything besides design and review.
That's like maybe 1% of total token usage for SWE BUT a better design can have enormous benefits, so it's almost always worth using the best model for that.
I'm not surprised at all that they might be doing shady shit, my free request was used up in less than 7 minutes and said pay to continue, no thanks. I've paid for monthly before, never again I don't care how good the model is. I asked about the extreme limited usage in their discord and got banned, not a good look. You don't have a way to measure what you're actually paying for and they can change it at any time.
yeah, we need to make sure we don't get stuck with these two evil corps. Dario is really the worst. Plays the drama queen but he is very hard in all things dark(i.e. dark patterns).
That viral Tweet was analyzing the summarizer output, not Opus output.
The reasoning traces aren't available. The summarizer gives a short summary of what the reasoning traces do.
The changes were in the summarizer model. The Tweet doesn't even say that Opus was nerfed, everyone just assumed it because the summarizer started using "I'm..." more.
This mystery was solved on Twitter over the weekend. Watching the pitchforks come out on Hacker News on Monday over a situation that was already resolved and debunked is starting to feel like how LinkedIn catches on to internet drama about a week later.
I don't have a formal benchmark, but daily work (coding mostly, plus some complex architecture design and review) has had consistent outputs today for me. No noticeable nerf from my perspective. I definitely took notice in some past instances, but if they nerfed Opus 5.5, then I guess it was just so good that I was nowhere near its limits.
In a normal world you are not allowed to degrade the service during your agreed and paid contract. You would get sued. Apparently this is perfectly fine in this current world.
According to our measurements this isn't true. We re-ran Opus 5.5 today on the benchmark we ran in September, with the same prompt and settings. Before and after, side by side: https://playcode.io/blog/macbook-svg-benchmark#reruns
Max, streamed (128K cap): 128,000 tokens, $2.56, empty file, in September and today.
Xhigh: 86,096 tokens for $1.72 in September, 107,333 for $2.15 today.
The Batch API bills half price. Max ran out of the 128K ceiling again, to the token. Given 300K, it drew the same kind of laptop with 5% fewer tokens. Xhigh used 25% more and drew the same kind of laptop. That's inside the noise: 17 identical calls at high ranged from 17,757 to 59,705 tokens.
The model is still the best of everything we've tried. Most likely the next best would be Fable 5.2. But we don't like Fable because essentially the result is the same, just the price is higher. Its medium and high are close to what Opus 5.5 Max had, but the price is higher.
(I run Playcode, an AI app builder. We benchmark every model before we add it.)
Is this voodoo? Every time someone makes this claim I'm left scratching my head wondering what the hell they're talking about. Either I'm not very perceptive, or these claims are subjective hand-waving.
A friend who doesn't follow these news suddenly told me he felt Claude was doing worse the past week. I bet him there was a new model version coming up within days. It's like clockwork at this point, no matter what deniers say. I got a beer out of it at least :)
It's been a mess over the last week or so. Went from being the best model I've ever used to utter dogshit that can't complete any task. It's very obvious with the Advisor pattern, where before when it chatted to Fable 5.1 there were small tweaks and so. No every time Fable comes back catching a ton of problems and bugs with Opus 5.5
stop OPUS 5.5 FUCK, DAMMIT, THEY DIDN’T JUST NERF HIM—THEY ABSOLUTELY OBLITERATED HIM, THE FUCKING MISANTHROPES! I EVEN MADE A MISTAKE IN MY ENGLISH IN THE PREVIOUS SENTENCE—NOW THAT FUCKING CLAUDE IS TEACHING ME ABOUT BUGS!
730 comments
[ 0.28 ms ] story [ 21.9 ms ] threadWas this before you lobotomized Opus 5.5 or after?
https://news.ycombinator.com/item?id=50419019
...but most of the benchmarks don't line up anymore because the benchmarks have changed.
The Humanity's Last Exam result matches the initial announcement, though.
Surprisingly the model is available immediately in my Claude Code: usually there's a delay, either for rollout or safety.
Fed to someone specifically to test if they'll leak it and to confirm who that is?
So, this is probably Fable 5.5; it's just named Fable 5.2.
Also, this is probably Model 3 from METR's report [^1]. They would name Model 4 Fable 5.5, once (and if) they decide to release it.
[^1]: https://metr.org/blog/2026-11-05-anthropic-ai-rd-acceleratio...
It might explain why DeepSeek Flash v4.1 1017 is the most popular model at OpenRouter.
by number of tokens? Also, we don't know numbers for native cli tools.
I would suggest calibrating what "a lot" means. In terms of dollars per unit of intelligence, it's extremely cost effective.
People don’t care what the costs for the LLM providers are. They are paying for a service. 30$ a month is reasonable for a lot of devs. 100$ is a lot for many. 200$ is unreasonable for personal usage for most.
Everybody understands the economics are all skewed and this tech requires a lot of resources / compute. There are still absolutes in terms of pricing that are not acceptable/affordable by most.
I'm open to learning more, not trying to ask a rhetorical question.
> the number of Fable resets you get depends on how long you’ve been a paid subscriber [...] Subscription length is measured from the date of your first paid bill. If you canceled and came back, the time since your first paid bill still counts.
> These resets expire on March 8, 2027 at 11:59 pm PT.
tl;dr if you joined when Opus 4.5 became a thing, you got 2 resets.
I couldn't use prior Claude models without rolling my eyes at the writing every few sentences, Astra was refreshing and then Opus 5.5 made even Astra look terrible in comparison.
Even the Plain English hack only is a temporary solution.
Interestingly, you can have informative conversations about the LLM linguistic tics with the LLMs themselves. One word that chatgpt used to describe what I was observing was linguistic anaphora (not rhetorical). It’s basically shortening/compression achieved via pronoun, and Claude does a lot of it. Worth reading about. A lot of it seems to do with how the new models are trained for brevity.
Another pattern I’m starting to become very sensitive to is Claude making up on the fly jargon for something under consideration, especially in code review. This morning for example it referred to “a claim” out of nowhere, and when I questioned it, it admitted that it was using the noun as shorthand for a record getting dequeued (“claimed”) inside of an sql transaction. It makes sense if you already know what it is referring to but in a code review it’s totally incomprehensible on first read, and ultimately counterproductive.
1. https://github.com/mattpocock/skills/blob/main/docs/engineer...
Opus 5.5 is a massive improvement.
You can also customize further by putting notes in your global MD.
I've had more success running a prose linter[1] with a ruleset[2] that gets rid of most of the egregious stuff. You can also pretty easily write custom rules for it.
And sometimes, when I'm fed up enough, I just throw all the prose in the codebase at Antigravity/Gemini to rephrase. The result is so much more readable.
[1]: https://vale.sh/
[2]: https://github.com/tbhb/vale-ai-tells
On the web with both high and xhigh, it had knowledge of both.
They're not typically known for their interpersonal skills.
Heck, some of the most common debates are on the semantic clarity of abstractions or syntax.
terra has been prettt good for me for me, when i need something looked at an explained to me.
i dont ask for much non-technical output though, which i expect yeah is its own cans of worms.
I prefer Gemini for day to day use as I find it a lot more agreeable to be working with all day every day, and Gemini is pretty much on a par for coding Vs Opus 5.5, at least for my needs.
I wonder why that team singles Gemini out like that.
Reference: https://x.com/METR_Evals/status/2118412205662314935
Pretty crazy that Anthropic already has another model waiting in the wings that is significantly better than this.
You are reading what the external testers wrote about the earlier checkpoint
Honestly incredible that they’ve managed to beat all benchmarks, while making it cheaper. Always happy to see improvements offered at lower costs.
Will be interesting to see what OpenAI does in response
i'm begging people to stop looking at token prices
> Fable 5.2 at High effort scored about level with Opus 5.5 at Max effort, for roughly half the cost per task
- https://news.ycombinator.com/item?id=50422069
Anthropic does this time and time again, unserious company
https://x.com/follistatindev/status/2119129611510383014
With early handouts of Qwen4 starting to be reported 5 days ago ... so if they're good, they would have had some hours to evaluate Qwen4 against Claude 5.5 and, then or right after was when Anthropic decided to put out 5.2 ...
So yeah ... benchmaxxing and all that. BUT nevertheless, the published results seem to say that Qwen4-Max appears to match Opus 5.5 on some benchmarks, and Qwen4-27B (probably by going even harder than Qwen3.8-27B on thinking/inference time compute) can beat Sonnet 5.5 on some things.
I guess they're not taking any chances. Keep in mind that Anthropic has declared they have non-public models (it's in their IPO filing). These are referred to by the shorthand names Model 2, Model 3, and Model 4. Essentially 4 is better than 3 is better than 2.
Fable 5.2 is "2.5" so to speak. So Anthropic was holding back, but not anymore. And now they can dial the performance up once more with what they have ready, and probably once more after that.
It would be funny though if you are right and that model is Fable 5.5 in a couple weeks.
so not everyone will have the same experience. If I was saving on costs that would be the easiest way to do it without deploying, without any nefarious intent. The persons in charge of that likely don't even know about how individual users are impacted
For example a provider quietly underclocks your cpu or puts it is an energy saving plan hoping you won't notice.
Or switch to a faster, cheaper cache mechanism or something that's almost as good, but much cheaper to run, in the name of improving efficiency and keeping up with demand. This switch to cheaper mechanisms is what it feels like to me.
Sometimes you just have insufficient evidence to be sure of something either way, and nobody gets to win the argument or shut somebody else down.
In this case, some have reported analytical work[0] that goes beyond "gut feel" anecodotal evidence, but even that's not going to conclusive and may not be trustworthy at all.
[0] https://www.reddit.com/r/ClaudeCode/comments/1xr9x5y/the_ner...
Ok, it probably burns tokens faster, but is it really a nerf?
They don't detect a quantized model until the quantization gets down to 4 bits or less.
Humans absolutely do detect quantization well before 4 bits...
Try it yourself if you want. Try a small local model, both quantized and not, and I would be very surprised if it isn't immediately obvious which is which.
But no, evals and benchmarks do not detect quantization, until around 4 bit quantization.
https://developers.redhat.com/articles/2024/10/17/we-ran-ove...
https://pub.towardsai.net/your-model-passed-safety-evals-the...
And the second article seems to reinforce the idea. Yes, broad perplexity tests are not a good proxy for the tests used during training.
Quantizing anything is effectively changing the weights, even if you don’t quantize the weights themselves.
https://news.ycombinator.com/item?id=50422951
So only if you were "selected" for their experiments, your account was nerfed.
If the nerfing is lower effort (seems to be, presumably why they hide the CoT) then they may be wondering if they can get away permanently with lowering effort on the new release provided it gives a wee improvement.
I’m working on a task. And the task is never to evaluate the agent performance.
If this turns out to be accurate, I think that would be a pretty unethical thing for Anthropic. Even if it is just a "we had to allocate compute to this new model release".
It should be fraud to swap out the model. We need a term for this, something like a bait-and-switch.
if they just straight up dont sell you something you bought then its fraud.
I think there is a ripe opportunity for expanding the definition of fraud to include changing the "ingredients" of a named product on the market. It is not limited here, we see it with food, and even in the medical field
(I will die on the hill that medical fraud was committed when mRNA therapy was used as a treatment for covid and it was 'sold' to the public as a "vaccine" despite not being the same thing as any vaccine before it. The definition of vaccine was expanded to include it without sifficient public disclosure to warrant informed consent imo. This is not a statement about conspiracy or the efficacy of vaccines or a moral judgement on whether it was for the greater good or not.. it was simply an act of fraud that robbed the public of proper informed consent by making people believe it was X when it was really Y.)
So yes, I think there is a large need for the public to at the very least put a strong handle on the ability to change what a product name means. It is a trivial problem to solve considering it is only a problem due fraud or fraud adjacent activities.
heres the solution: You have your product name, and then any changes to it REQUIRE a new version name. Like how you have Coke and Diet Coke, only formalized such that Coke cannot legally replace the ingredients of regular Coke with the ingredients of Diet coke and still call it Coke. With Software they could at least use version numbers and release modifiers. most of the time its built in already, but it needs to include the full stack. if they change anything that has an interaction with the user it should change the label seen by the user for the product.
thank you for coming to my ted talk
A Opus 5.5 is not a Opus 5.5. Its Opus 5.5 of some checkpoint.
mRNA btw. was well known how it worked, it wasn't hidden and some people waited for the non mRNA version because of it. I never seen or read the argument you make and I do not get it.
A vaccine is something which protects you from an infection. What definition do you think changed in humans? That we make the protein? How is that so different from viral vector vaccines?
The consent is a certain level of trust from the society to a group of researchers limited by being humans and having to act in a certain time frame.
Nontheless, I do agree I would like to have a fingerprint and proper changelogs etc. for products I buy including digital and analog. But this has a downside as usual: The smaller you are, the more it costs you to (probably) register, maintain and change.
You know the good thing about Claude etc. right now? It doesn't matter. My invest is not 4 digits and there are that many changes, i'm quite happy to have been upgraded for 'free' from Opus 4.8 to Opus 5 (what do you call this upgrade, which was significant better?)
I'm a bit disappointed with how this nonsense is spread across HN today.
Are you saying that the many, many people posting and experiencing this are wrong and you are right?
And not just in this instance, nerfing has been complained about in a cyclical fashion for many months.
But, they're all wrong?
You're so confident in your experience that you cannot think that other's experiences are also true?
"Could all these people seriously be wrong?" Yes, yes they could.
No, it gets complained about constantly even if there is no model coming out. A completely unrelated thread on Claude will feature comments like "can I just say that [model] has suddenly become useless this week?" Merely asking that question is bound to get people agreeing with you because we are evaluating all of this on vibes and placebos.
I too feel that way sometimes, I don't think it's ridiculous that Anthropic might silently downgrade you as demand warrants. But asking others if they "feel" something I feel is less than useless.
Like most experiments do not save the company money, there was an experiment to test the new fable, people got to use it at opus-5.5 usage. But this experiment clears the thinking blocks when people go back to the vanilla model; not sure if Anthropic overlooked this, but idk, it does not sound like "nerf on purpose".
There are experiments to adjust the effort level, at the moment the ones that I've seen switch to a higher effort. Again this is not a nerf; might be annoying perhaps, the model does start to behave differently (and consume more tokens). But it's not a nerf.
Then there is apparently one experiment that replaces opus-5.5 system prompt with the one from opus-4.8; the model is the same; this experiment is perhaps a bit more questionable, but still it's not "anthropic nerfed the models".
If people think they are getting nerfed they should check what experiments are enabled (claude.json and session logs).
The media coined the term "dual quality" for a European food scandal a few years back, in which conglomerates (including Coca-Cola) adulterated products with cheap, substitute ingredients, if they were destined for distribution in specific markets (mainly, lower-income).
The packaging exteriors were the same; and there were (I think)? broadly no price differences. It was essentially what the software industry calls "A/B testing", except with sneaking things into what people eat.
I don't have any good reading links on hand; here's one source,
https://ec.europa.eu/commission/presscorner/detail/en/qanda_... ("Dual Food Quality: Questions and Answers")
Right?
Eval context: 30 bug fixes and features evaluated by tests the agent didn't see. Ran on Opus 5.5 at medium effort in Claude Code.
20 of the tasks were last run on Nov 1, twice each. Nov 1 solved 36/40, and today solved 19/20. Thinking tokens per task were 0.95x Nov 1’s and output tokens 0.97x. The one significant change was speed: median time to first token decreased from 3.0s to 1.7s.
Caveats: one run per task today, medium effort only.
Paragraph 1 says there were 30 bug fixes.
Paragraph 2 first says they tested it on 20 of the “tasks”. But why is the denominator 40? Where did that come from? It should be 30.
It goes on to mention another test, this time with 20 tasks. Again, it’s unclear where this is even coming from. Based on the first paragraph it should be 30 not 20.
Why paragraph 1 even mentions the full set of 30 tasks if 10 of them are irrelevant to the comparison, I don't know.
I'm not sure if it's LLM text, or somebody who has been reading way too much LLM text and is now corrupted into mimicking it. My guess would be a mix of human and LLM. "...evaluated by tests the agent didn't see" sounds very Claude. The first line (2 sentences) sounds human to me. Everything after sounds AI, though the last line is ambiguous.
It's kind of sad how our meat brains are getting non-consensually trained on this categorization task....
2 alternative interpretations to consider:
1. 30 was a typo, it should have been 20.
2. 30 was the full suite, but they only ran 20 of those 30 on Nov 1st and so ran the same corpus again for a fair comparison.
> Paragraph 2 first says they tested it on 20 of the “tasks”. But why is the denominator 40? Where did that come from? It should be 30.
This one is on you:
> 20 of the tasks were last run on Nov 1, twice each.
There are 30 tasks in my eval total:
- 20 of them ran on Nov 1, twice each (so 40)
- I created 10 more tasks and ran them on Nov 5 (also ran them twice each)
- Today I ran all 30 tasks, once each, on Opus 5.5 to see if the performance decreased at all
- I only put the comparison against the original 20 in my post, since I saw in another post that people are claiming Opus was nerfed on November 3
This will all get less confusing when I make my benchmark public (hopefully this week).
Also - LLM text? If GP is confusing it's because it's overly succinct and might have typos, which are not exactly the hallmarks of LLM text
System prompt would be the easiest one.
But there were flash or inbetween models, different checkpoints etc. THey might also try different hardware which might not be binary compatible.
Opus 5.5 was so good that if it wasn't nerfed - at least just looking at the benchmarks - the difference in cost here does not seem to justify using this for anything besides design and review.
That's like maybe 1% of total token usage for SWE BUT a better design can have enormous benefits, so it's almost always worth using the best model for that.
The reasoning traces aren't available. The summarizer gives a short summary of what the reasoning traces do.
The changes were in the summarizer model. The Tweet doesn't even say that Opus was nerfed, everyone just assumed it because the summarizer started using "I'm..." more.
This mystery was solved on Twitter over the weekend. Watching the pitchforks come out on Hacker News on Monday over a situation that was already resolved and debunked is starting to feel like how LinkedIn catches on to internet drama about a week later.
Fable 5.1 Max $11.46
Fable 5.2 Max $44.70
Max, streamed (128K cap): 128,000 tokens, $2.56, empty file, in September and today.
Max, Batch API (300K): 255,957 tokens for $2.56 in September, 243,707 for $2.44 today. The two drawings: https://playcode.io/marketing/assets/static/claude-opus-5-5-... and https://playcode.io/marketing/assets/static/claude-opus-5-5-...
Xhigh: 86,096 tokens for $1.72 in September, 107,333 for $2.15 today.
The Batch API bills half price. Max ran out of the 128K ceiling again, to the token. Given 300K, it drew the same kind of laptop with 5% fewer tokens. Xhigh used 25% more and drew the same kind of laptop. That's inside the noise: 17 identical calls at high ranged from 17,757 to 59,705 tokens.
The model is still the best of everything we've tried. Most likely the next best would be Fable 5.2. But we don't like Fable because essentially the result is the same, just the price is higher. Its medium and high are close to what Opus 5.5 Max had, but the price is higher.
(I run Playcode, an AI app builder. We benchmark every model before we add it.)
Fuck Anthropic.