I should find information about the user's concern instead of just assuming.
The outage is a real concern, and the issue is worse than we realized. Requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 encountered elevated error rates. Worth stating plainly: these are not just models — they are load bearing rungs on the software development tooling ladder, and a blocker on this level makes the outage really bite.
One decision that is yours to make, not mine: should an email be drafted to Anthropic support? This issue has teeth, and a canonical handoff can land us where the main gate is no longer breaking silently.
If you're at the top of the screen, at least in Chrome 153.0.8010.37, it has a little interactive bit. You have to scroll through the images in order to be dropped at the actual web page, at which point the images go back to being a regular part of the page.
It's less than OpenAI did for Astra, but that was my first encounter opening it and my first thought was that they decided they liked Astra's hero/scrolling animation. I'm pleased to see they didn't make the entire page that like OpenAI did, but I'm expecting to encounter this pattern more often on these announcements now.
Hijacking the scroll wheel has existing long before "AI". Many "high end design" websites that want to "tell a story" get woo'd into thinking it's a good idea. It's terrible, and feels like your scroll wheel is stuck in quicksand.
didn't they say Opus 5 was Fable-level too tho? Let's see, I'm at the point where I don't think benchmarks really tell us very much any more. I'd love it to be as strong as Fable, but I'm skeptical about how that will look in practice.
Opus 5.5 communicates more naturally than prior models. Early testers found its writing clearer and easier to follow, which addresses some of the common feedback we heard about Opus 5.
Sounds like they noticed the complaints. I'm curious to see what LLM-isms this one may have.
It was difficult to not notice them. Opus 5 was unusable, most of my team went back to Opus 4.6 for most of their work. I hope we can move forward now.
How? Explicit instructions, memories and even skills have not been able to keep Claude from saying "genuinely" every two sentences and keep it from explaining heavily what something _isn't_.
I appreciate humor here, but there are now a dozen of these comments on every thread about Claude. They no longer adding anything substantial and dilute the discussion.
I don't mean to pick on this comment in particular. The majority of my work day is now spent reading AI generated text, and I look at HN (too much!) because I want to read human commentary. Humans pretending to be obnoxious AI on repeat is net negative to say the least.
I agree. Hopefully Anthropic has fixed Opus' ridiculous communication style so that people - like me - no longer have any kind of weird impulse to imitate it.
I'm genuinely confused what's the relationship between LLMs improvements and them being so incoherent.
and it's not about the verboseness (even though it obviously contributes to the fatigue and loss of focus), I swear the vocabulary of the llms change working on the same task on the same codebase significantly.
Can it be that now they are getting optimized against benchmarks that are valuing logics, rather than human appreciation? (I am not an expert at all, just an idea)
Small quirks can quickly add up in posttraining if not caught. Although TBH with how obvious Claude language is, I do feel like this is something Anthropic probably noticed and just assumed people would not care about. Now that people have obviously cared, they're probably actively looking to alleviate it
Opus 5 was just incoherent - curious to see what improvements they have made here. Would love to see some kind of postmortem to better understand how writing styles change from model to model.
I wouldn’t be surprised if Opus 5 was trained on content written by other LLMs
I don't think it's substantially different. I just pasted a random chunk of code and asked Opus 5.5 to comment on it:
The Vercel target is hard-coded. That's common and not wrong, but it's opaque; nobody reading this later will know which Vercel project it belongs to, and if the project is recreated the target changes silently. A comment or a named variable would help.
Pointing a DNS name at Vercel is only half the job. The domain also has to be added to the project in Vercel's dashboard, otherwise requests will arrive and Vercel will reject them. That step lives outside this code, so it's easy to forget.
Finally, [CENSORED] existing only in production is slightly odd on the face of it. It may be perfectly deliberate
(perhaps a single shared testing tool that only needs one public address), but if you're reviewing this rather than just reading it, that's worth confirming.
It has the same annoying cadence and writing style with slightly less prominent claudisms.
Maybe if we had a single human we talk to 24/7 at scale, we would get annoyed at his cadence and style. You need variety to not pick up on known patterns I assume, which a single model can’t replicate?
No, it's just poor writing. Actionable points are buried inside the paragraphs and over-hedged, and one point is completely made up. Compare to a five second rewrite:
* Consider leaving a comment about the hard-coded Vercel target. It's not clear where does it come from.
* [This is just a bullshit point, because the domain is not "added to" Vercel, it's provided by Vercel]
* Are you sure that [CENSORED] is prod-only? The name suggests otherwise. [also, what "if you're reviewing this rather than just reading it" even means?]
also, what "if you're reviewing this rather than just reading it" even means?
It means "I'm treating you as lay-person punter, not a developer working on this project." Opus 5 feels like it's constantly trying to reward-hack me into treating it as intellectually honest and epistemically humble, while in the same breath it talks down to me and tries to smuggle its own bullshit assumptions and assertions into the conversation unchallenged. No progress on this front apparently. Glad I cancelled.
I did that but I recognize that even though your submission was a few minutes later than that one, you posted the better link, and you're also an established account (the other post was from a new/throwaway account), so I've restored this submission and moved the comments back to it to reward you.
Which is one of those fun things that didn’t actually exist back when we took it for granted that our fellow person was operating under some kind of moral or ethical framework, which pretty much everyone was until the economists told us that wasn’t rational, because it turns out it’s an evolutionary advantage to operate under an ethical or moral framework because it allows the kind of coordination which facilitates better collective outcomes, which everyone knew until the economists came along to tell us we were wrong and in fact it was rational not to do so and suddenly we had the prisoner’s dilemma.
And how was the Xbox 360 naming choice a “debacle”, exactly?
It was odd at the time, yes, but no one really minded it truly. Heck, Xbox “ONE” was a lot more of a fiasco/debacle than “360”—but there’s no parallels to be drawn with “ONE” here.
I see what you’re trying to get at with this comparison, but a “debacle” it ain’t.
As long as it's not as verbose as Opus 5, I am quite happy with a better version that's also less expensive. I will test it tonight. Grok 4.7 was horrible, and for mundane tasks I am relying on DeepSeek Flash 4.1 with great success using OpenCode.
my projection is that they are still gonna be pretty far behind, but they will sew it up in the next few releases. it feels like they were caught with their pants down on how much work OpenAI has put into that area, but i doubt there is some magical secret sauce that OpenAI has that Anthropic simply cannot catch up with.
Interestingly they've changed their approach to usage resets for this release - with previous releases I've had my usage instantly reset, but now in the Claude app I've got a 'Reset for free' button that expires Oct 22, which seems to effectively be a whole new usage window I can activate whenever's convenient
We see signs that Opus 5.5 often suspects it is being evaluated, which challenges our ability to assess how it will act in the vast variety of real-world settings it is deployed in.
We can't test it properly because it knows it's being tested.
On our benchmarks, Claude Opus 5.5 leads in agentic coding, computer use, and knowledge work. That said, at these levels of capability we’ve found that benchmark margins have become a less reliable guide to real-world differences. In our own use, the gap between Opus 5.5 and Claude Fable 5.1 is narrower than these scores suggest.
So it beats Fable 5.1, by quite a bit, on every metric? Interesting.
Might have to use my $20 Claude sub some more. I was moving away from it to a $100 OpenAI one to avoid the Claudese and poor token efficiency of Opus 5, given that I couldn't use Fable 5.1 with my tier, but this is worth trying out.
Why can't they let 20usd claude subscriptions access fable in CC, as openai allows you to use astra and max modes in codex - you just pay for it in more token use.
It performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.
Input and output tokens are $4 and $20 per million, 20% less than Opus 5. Cache reads (which make up the majority of agentic and coding work costs) are $0.20 per million tokens, 60% less than Opus 5. Opus 5.5 also generates output more than 30% faster than Opus 5.
Better than Fable, cheaper than even the last Opus. I use Opus as my main driver so this is very exciting!
Not the same person but... nothing. Haiku just hasn't been an interesting model for a long time. If you want cheap and fast, there are lots of options that are simultaneously cheaper, faster, and capable than Haiku.
If you're able to use the OpenAI ecosystem, Luna's price/performance is really good. Almost like "they messed up and accidentally made it too good" good.
We use Haiku 4.5 inside our product. It continues to be absurdly capable for converting natural language to structured JSON based on a set of fairly complex business rules.
They mention "the first model in our new Claude 5.5 family". Obviously that means Fable 5.5, but hopefully also a usable update to Sonnet and Haiku. Sonnet 5 hasn't really had a place in the line up for anyone I feel.
Maybe Anthropic finally felt the pressure from MiMo, DeepSeek, GLM Flash and Luna.
I thought about taking a shot every time Opus 5 said "load bearing", "bites", "teeth" (real oral fixation it had), "real {concern,issue,problem,...}" and realized I'd be dead of acute alcohol poisoning by lunch if I did so.
The problem with 5 wasn't just the verbosity, but its insane way of communicating. It had this bizarre circuitous sentence structure that always buried the lede, and always tried to be faux profound. I'm okay with verbosity if it's actually readable.
It is hilarious to me that in the examples they show side by side Opus 5.5 still uses 4 times more words than it needs to use.
IME, if you eyeball how many words the thing they're trying to say actually needs, and tell them to use only this many words, they become excellent communicators. I assume something about Anthropic's grader for writing just really wants to tick all its tidy tiny boxes of information the models need to cite. It's terrible.
Astra is just so good. And the ChatGPT subscription lets me use my own harness, so I can hook it up to exe.dev.
I’ll be going about my day, have a random idea, launch a microvm on exe.dev with a prompt of my idea, and get a working thing a few minutes later.
I don’t know how much better a model would have to be to get me to move off OpenAI at this point, but doing just a little bit better in terminal bench 4 isn’t it. It would have to be a difference in kind, like opening up the harness restrictions, or privacy guarantees (comparable to offline models).
Edit to address questions below:
ChatGPT supports oauth login.
Exe.dev has it built in. IIRC, pi also has it built in via /login.
ya i've been a gpt hater for a while. almost exclusively used claude up until astra. astra feels like it blows everything out of the water. its fast, correct, organized, and less verbose.
Anthropic is absurdly vague about 3rd party harnesses for subscriptions, if you try to use anything besides Claude Code, you are likely at risk of getting banned, you can "do it", but are at their mercy if they decide to ban you. OpenAI gives their blessing to using oauth on any harness, you can make your own or use any of the popular public ones like opencode, pi, whatever exe.dev is that this guy mentioned.
So in simple terms, OpenAI doesn't restrict you to Codex, and gives their blessing to try whatever you want with their models(besides serving others with your subscription usage, that is still afaik against tos).
I use the other 50% of my $200/mo Claude subscription by having Fable run Opus subagents for a lot of work. That way I don't have to deal with Opus directly.
I think this release is really going to give them a hard time selling Fable:
On our benchmarks, Claude Opus 5.5 leads in agentic coding, computer use, and knowledge work. That said, at these levels of capability we’ve found that benchmark margins have become a less reliable guide to real-world differences. In our own use, the gap between Opus 5.5 and Claude Fable 5.1 is narrower than these scores suggest.
In general, "benchmark margins have become a less reliable guide to real-world differences" sounds like a big problem. It was certainly the biggest problem with the previous generation of Claude models for a different reason, because the non-code output was nonsensical, and that is not being benchmarked at the moment. But I'm not sure what to make of this admission.
Opus 5 fucking sucks. Like it's horrible. I use Fable for coding and anything important and I use Opus 4.8 for things like recursive email categorization, transaction matching, and other stuff where I don't want to burn as much quota.
In my experience Opus 5 is the worst of all possible worlds, it's dumb and headstrong. It just runs away with tasks you didn't ask it to do, is reckless, and basically is unusable in my experience.
Not sure why but my guess is that this will be worse. Happy to be proven wrong.
I believe Opus 5 isn't meant to be spoken to by humans. It's great at executing but I reckon it's intended to be spoken to by other models such as Fable. I use Fable as the orchestrator, only speak with Fable, and all implementation, recon, design etc happens with Opus 5, with Fable reviewing (and translating).
I've really gone in the opposite direction: having a dumber model orchestrate. In my case, it's usually a Luna orchestrator spawning Sol/Astra subagents to do the "big brain" work of planning and reviewing.
Reason I went with "dumb orchestrator" was just to save tokens. Having Opus/Sol (let alone Fable/Astra) orchestrate was burning tokens like crazy for me even when much of the gruntwork was being done by Luna/Sonnet/Haiku subagents. (Luna is also really good, like way better than Sonnet...) Perhaps it was a skill issue on my end though, maybe I wasn't just managing context properly.
Yeah if anything Opus 5 taught me how little benchmarks mean to the actual real world performance of these models.
"Better" in every sense of the benchmarks and absolutely horrible results in my day-to-day work.
The verbosity, goal post moving, tendency to leave work unfinished, over focusing on unrealistic root causes when debugging, etc... etc...
It was the first time I actually pinned my models back because I just could not work with 5 for the price and performance it gave me. Hoping 5.5 is better this time around....
"benchmark margins have become a less
reliable guide to real-world differences"
sounds like a big problem.
My guesses:
1. Real-world use cases typically involve big, hairy, crufty, tech debt laden codebases and benchmarks do not.
2. AFAIK "success" in a benchmark essentially boils down to "do the tests pass and do we get the right result?" which is something the LLMs have been achieving with ease for a while, except maybe for uber-challenging coding tasks that would be outliers in just about any workplace. Whereas real-world software engineering is usually just a bunch of CRUD... and "success" involves harder to measure dimensions like "maintainability" and "did you overengineer this?" and "how did you cope with a bunch of vague and maybe contradictory business requirements?"
Having said all of that, I have never ever looked inside any of these benchmarks. I'm putting my guesses out here strictly in the tradition of "the quickest way to learn about something is to be wrong about it on the internet."
Prices per 1M tokens Claude Opus 5.5 Claude Opus 5
Cache reads $0.20 $0.50
Input tokens $4 $5
Output tokens $20 $25
Cache writes $5 $6.25
Opus 5 is the model with highest spend on openrouter (https://openrouter.ai/rankings#task-spend) and it seems plausible that Opus 5 is/was the highest spend model in the world, and certainly Anthropic's biggest moneymaker.
If you are forced to reduce price despite raising capabilities, that certainly tells something about the market, and potentially about Anthropic future profitability too, since this model is their biggest topline contributor
I think gpt 5.6 family also dropped pricing but didn't give any more usage for the subscriptions. Maybe it's a way to silently lower the value given to subscriptions while keeping API pricing competitive
We haven’t been able to use opus as much as we’d want because it’s been too expensive for general use, price drop is good so I can stop juggling different models and just use this daily unless it has some weird new issues
It’s good at finding and fixing inefficiencies in software
Holy shit! Its happening!
Now if we can the AI to understand this *implicitly* so that it doesn't need to be stated upfront, we might be able to undo years of "premature optimization is the root of all evil".
Because Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, we’re deploying it with safeguards similar to those on Claude Fable 5.1. Vetted organizations can apply today to our Life Sciences Verification Program to use Opus 5.5 for biology research. In the coming weeks we will also be expanding access to our Cyber Verification Program, and verified cybersecurity practitioners will be able to use Opus 5.5 for their work.
Ah, they're spreading their limits to all their models it seems. Definitely not a good thing long term in my opinion.
It’s not like ChatGPT isn’t doing similar. I’ve been hit by cybersecurity strikes before while working on an internal codebase that I had to appeal. Anthropic hasn’t done that to me yet. ChatGPT also regularly does that “thinking for a long time while we check if your chat is rule breaking” thing a lot for me when doing model identification without even interacting with external codebases or services.
The real answer is local instantiations where you don’t have to worry about poorly tuned guardrails screwing you over while you try to work.
Until eventually the Chinese models get good enough/the strategic balance shifts and they start locking everything behind closed weights the same way the US companies are doing.
For some cybersecurity tasks, the Chinese models are already good enough, things like PoC development or things like exploiting mis-configurations.
Whilst I'm sure the top-end OpenAI/Anthropic models might be better, I've found their guardrails so twitchy (especially Anthropic) that I wouldn't try to use them for even vaguely security related work.
One of my favorite things about their safeguards is their own model will utter something which it does not like and then I'll need to reset the conversation.
The safeguards really don't work well for a lot of long-running tasks on old code bases. A lot of my workloads last days to weeks and the single biggest risk to the workflow is random safeguards.
Many people knew this announcement was coming. The betting markets suggested a very high likelihood of Opus dropping today. I was anticipating this quite a bit!
With the performance gains they're claiming, I wonder if they implemented the Casual Encoder-Decoder technology from DeepSeek 4.1's paper.
I could see them accomplishing it and seeing gains like this in roughly the correct timeframe, and when I heard about that development I assumed the frontiers would probably jump on it.
Claude Opus 5.5 is our first release since we called for pacing the frontier.
Interesting how the very first line is used to remind the reader of their call to pace the frontier just last week, and everything else after that line is to demonstrate with very specific numbers how they absolutely are not pacing.
It seems like they are. I mean, this is similar in performance to Fable (ish). It seems like more focus on making existing capabilities more accessible.
Everyone knows both labs have internal models which outperform the frontier. All releases are to match market parity and demand for spend, the rest of the compute is used for training. It's not worth arguing about this.
A razor is a philosophical tool to help decide between options, in the case of Occam it’s a way to decide for something in a situation where multiple options have more or less the same level of plausibility to en your current knowledge. It’s a heuristic to make a “cut”. What are you shaving off?
Maybe not replaced exactly but they won't be manually typing out lines of code anymore. I haven't written a line of code in like 6 months. I review PRs, write prompts and tickets, check CI output, and get frustrated when the magical code machine stops working or I run over token budget
Dario never said they would be. Just that more and more code will be produced by LLMs. All his predictions were in fact pretty much right in terms of months and percentages, give or take small margins.
Fable 5.1 came out just 21 days ago. Only 3 weeks! And this is 20% relative improvement on terminal bench vs Fable 5.1 at less than half the price, and more human sounding output. does not feel paced to me tbh.
I guess I understand why we'd want to flatten a COVID curve, but why do people want to flatten the LLM development curve? Don't we want the opposite? Isn't the goal AGI?
Well, there's a tension because, depending on who you ask, AGI is how you cure cancer and achieve utopia, but also how you kill all life on earth and turn the solar system into paperclips
There is a difference between wanting AGI (which not everyone does), and wanting it as fast as possible no matter the side effects and potential for vast harm.
Homo sapiens is 300k years old, maybe it’s ok to delay AGI by like… 1 year if it meaningfully improve our ability to align ?
What insights do you have? Because the blog/benchmarks don't imply any "ish" ... this is crushing Fable across the board. The only nuance to this is how the employees are saying what you're saying: "similar performance" but again what does this mean vs what is presented to us?
You mean the section where they tell us this model is not affected by pacing because “they understand it well” and they will share more details on pacing later? Yea not very convinced by this effort.
It's laughable at this point. It feels like they're drumming up all this fear about imminent AI threats to emphasize the need to slow down, when in reality, the model progress seems already to be slowing down and has shifted to compute allocation (i.e. "how much compute do you want to throw at this prompt?"). All while continuing to tout benchmark records with each new release.
Either AI will capture so much value that there will be permanent underclass (and in this case it's extremely capable and extremely dangerous and should be heavily regulated) or it won't be capable enough to displace people into permanent underclass.
I really don't think it's productive for internet forums to constantly be criticizing language choice when the meaning is clear. Better to respond to the substance of the issue than word choice.
Is the meaning clear? Nobody would use "pacing" in this way. I only know what it means because I've seen previous press releases; if someone told me they wanted to pace the frontier I would have no idea what they mean.
I'm on the fence about calling out AI-isms but I think it's definitely worthwhile to call out ones that actually don't make sense.
It makes sense to me. If you 'pace your running', you're setting the speed intentionally. The phrasing doesn't describe whether it is fast pace or a slow pace, but it describes having a goal and not just winging it.
Granted I am not a native English speaker but I have no idea what "pace the frontier" means. When I read it I just assumed "frontier" refers to "top of models" and "pacing" is that they are getting there quick.
Is this the meaning or do I have it wrong? I have not checked.
Yeah, when I first saw it referenced I assumed it was from something Dario wrote previously and was now disavowing, meaning "keeping up with the frontier" (i.e. racing forward from behind to match pace). Like from back when Anthropic was founded to promise they'd quickly catch up with OpenAI or something.
The meaning was immediately obvious to me as a non-native speaker, and it sounds quite poetic. Pace makes complete sense in that this is perceived as a race, and 'the frontier' is pretty much the shortest, clearest way to say 'state of the art development of AI'.
The meaning isn't clear at all. So open for interpretation that it is meaningless. That's the whole fucking point. For all I know they are "pacing the frontier", or not. The fact that there's no meaning to it let's you know that it was a pointless waste of tokens and attention.
What specifically is smarmy and weird? Sounds like your own hot take with zero analysis.
They’re limiting frontier model development speed. Others are too. Pacing is the only word here to criticize, and I think it’s fine given the limiting of speed but also increased oversight. I’m not saying they’re fully doing this, but the term is fine.
Agreed when I first heard the phrase it sounded odd. Maybe they thought it subtly conveyed they would be setting the pace... But again this is something AI would come up with in its awkwardly post hoc sort of way.
Though tbf corporate-speak and AI-slop are both insufferable in similar ways...
They are advertising the regulations they want to enforce in the following sentence, which makes their intentions explicit (ie apply those things made to suit us to our competitors).
When they talk about pacing, they are referring to their dangerous competitors, particularly those evil open-sores and Chinese ones, not their lovely safe models because you can trust them to look after your interests.
What the big players are trying with the current calls to slow things down, is the standard capitalism practise of trying to engineer regulatory capture. TBH I'm surprised those calls are coming so soon - they must be really worried about running out of what little moat that they have.
For users with cybersecurity use cases that may be blocked by our cyber safeguards, we
recommend accessing our models with reduced cyber blocking classifiers via our Cyber
Verification Program. Claude Opus 5.5 will be available through this program in the near
future.
Anthropic has used "in the near future" for Mythos-class models too, but CVP is still Opus 5 only.
Why even have the program designed for trusted access to cyber capabilities if you're not providing access to cyber capable models via the program?
Because Opus 5.5 is comparable to Claude Mythos 5.1 in biology and cybersecurity, we’re deploying it with safeguards similar to those on Claude Fable 5.1
Great so good luck using this for any low-level embedded or operating system development (unless you really, really like Opus 4.8 and want to be greeted by its familiar face after a few minutes of work!)
Found this announcement interesting since allegedly OpenAI is retiring their Terra tier. I think for everyday work, two models with various thinking efforts seem enough, plus some frontier level model like Fable or Astra to coordinate.
Great that they listened! The improvement in communication style looks fantastic. Opus 5 was insufferable and I was on the verge of cancelling my subscription.
Distillation attacks, in which attackers use thousands of fake accounts to extract a model’s capabilities at industrial scale, create safety and national security risks. Distillation allows bad actors to create highly capable models without the safeguards we build into Claude.
Maybe its a bit tiresome to read another comment of the form "what about your large scale distillation attack on the Internet", but this statement really just pisses me off. How very insincere in the most aggravating way.
"bad actors" boogeyman, and we should trust some tech weasel to do the right thing? Yea we've seen who they really are, once they get a sliver of power.
Current models, especially Opus are almost unusable because they don’t respect instructions and their responses are infuriating. They are clearly designed for token consumption. I find myself wasting a lot of time just asking it to shorten or simplify its responses. I’ll give this new model a go, but I’m not holding my breath because the last model release was supposed to fix the very same issues and it didn’t.
We’ve made major improvements to the way Opus 5.5 writes and communicates, one of the most common areas of feedback we heard about Opus 5. Its messages are much easier to understand at a glance, which testers said helped during long working sessions.
Don't forget that Opus 5 was tracking fable on many benchmarks, yet it was borderline unusable for any coding work. My Claude sub usage has been 100% fable, 0% opus 5.
Benchmarks often don't survive contact with reality.
Mines pretty much inverted - my colleagues and I noticed almost zero difference between the quality of code in Opus vs Fable. Occasionally I'll switch to Fable for an arduous debugging task but that's about it.
That's not my experience at all. Opus is an extremely capable coder on high or xhigh effort. It can read academic papers, implement algorithms from the description in the paper alone and reproduce results without breaking a sweat. This is remarkable because it is pure reasoning on unseen material; in some cases the paper was just published and there wasn't an implementation to learn from in the training data.
Big model smell is a real thing. For certain classes of problem, ones you get a feel for but can't easily articulate, a big last-gen model can get you what you're looking for when no quantity of tokens from some ultra-RLed mid-size latest generation model can.
If you look at the kinds of tasks included in benchmarks, you get a feel for what’s likely to have been benchmaxxed and therefore what’s effectively in- or out-of-distribution for a model, and where a larger model's broader world knowledge is beneficial.
Eg SWE-bench used to include a lot of Django tasks, so smaller models got good at Django pretty quickly.
People aren't the target audience of that part of the post. They're hoping saying enough safety stuff will ward off the looming regulatory sledgehammer.
Anthropic is one of the most valuable companies in the world. Their comms are designed to appeal to a very wide readership.
HN is a bubble that's mostly out of touch with what regular people use or care about.
In 2007, HN was convinced that nobody uses Microsoft products. In 2016, it was that Facebook doesn't have any real users and is dying. In 2026, it seems like nobody cares about AI safety and everybody wants to run local models.
They purport 40% drop in costs due to lower token pricing (presumably aimed at winning back the many of us that switched providers in discovering Opus 5 unusable) and improved token efficiency.
"Early testers found its writing clearer and easier to follow, which addresses some of the common feedback we heard about Opus 5"
and
"We’ve made major improvements to the way Opus 5.5 writes and communicates, one of the most common areas of feedback we heard about Opus 5."
and
"In our own use, this has made Opus 5.5’s work easier to follow and check—which is a safety benefit as well as a practical one."
I realize it is corporate communications but "most common areas of feedback" and is a bit sterile. If the company wants authenticity and trust its easy to say that they found it hard to follow. And that it did not meet a quality bar they generally expect from their releases.
If this is not true, that it Opus 5 output was generally acceptable and we might see something like that again, that is an important consideration for potential customers or investors.
Distillation attacks, in which attackers use thousands of fake accounts to extract a model’s capabilities at industrial scale, create safety and national security risks. Distillation allows bad actors to create highly capable models without the safeguards we build into Claude. Our September 2026 threat intelligence report details the illicit distillation activity we’ve detected and disrupted so far.
Such a negative tone they put on this. Distillation is amazing, because it means anthropic and openai fail to keep a monopoly. Who even are they who claim it's unethical? If it is truly unethical, then so is the mass data scraping they do on my personal website on a regular basis (without my consent), and all the unauthorized use of content produced by authors, blog writers, wikipedia contributors, and creators everywhere. If it is truly unethical, then anthropic, openai, meta, google... all these companies should have deleted their LLMs long ago. This wording disgusts me.
Heck, it would be amazing if we had more models without guardrails - some of the models that are produced via heretic[1] are actually quite nice to use - in particular, I've enjoyed investigating Chinese censorship by interacting with an abliterated model of Qwen3.8-27b. If security is really a concern, then secure your systems - don't attempt to dumb-down the tools we use. If someone breaks your window, then they are responsible, not the hammer they use to do so.
I'm confused how they have been able to create so much public negative perception around distillation. It seems pretty clear that they are the only ones who lose out, and everyone else benefits. I don't have any ethical issues with it, nor is it illegal: at worst it's a ToS violation.
IMO the biggest problem with distillation is that not enough people are openly doing it. I would love to see more small, competitive US labs instead of having the eggs in 2~4 baskets (depending on how you count).
It crushes Fable on benchmarks and even in the blogs "real-world" studies. But... they are communicating like it ~sometimes~ provides Fable intelligence?
A bit confusing, otherwise I would assume this is a complete replacement for Fable across the board??
I have just switched to 5.5. First mistake was stale environment variable, didn't realize it was replaced, "oh my memory had stall data" and that's it. Second one, a powershell command had the wrong syntax. Great for my first two prompts.
I'm glad they specifically called out the prose issue, I was always pinned to Fable 5.1 because I wanted to avoid the unreadableness of other Anthropic models.
It performs at the level of Claude Fable 5.1 on most work
Fable feels a bit dated after using Astra. Progress is nice (just like Grok 4.7) but I'm looking forward to the next big release as I can't fully commit to Astra atm.
In the coming weeks we will also be expanding access to our Cyber Verification Program, and verified cybersecurity practitioners will be able to use Opus 5.5 for their work.
I was accepted into the CVP a little while ago. Does this mean I'll need to apply again?
> On our benchmarks, Claude Opus 5.5 leads in agentic coding, computer use, and knowledge work. That said, at these levels of capability we’ve found that benchmark margins have become a less reliable guide to real-world differences. In our own use, the gap between Opus 5.5 and Claude Fable 5.1 is narrower than these scores suggest.
Finally confirmation that Haiku was not forgotten and will be coming soon, althouhg I find it quite interesting they skipped 5 and directly skip to 5.5 with all models, including Sonnet which is not super old. I suspect they found something breaking that allows to release this. Recently they struggled with keeping up a 50 % weekly limit increase and now they're putting out 30-40% faster and cheaper models even faster, with much more better benchmarks, a limt reset command and five hour limit increase. It seems more like the opposite and as if they never struggled, thus, I very much believe they found something very effective and new.
I'll have to try 5.5 on my work's Cursor account. If they really solved the communication issues, I might consider moving my personal account from Codex back to Claude Code.
Communication. Opus 5.5 communicates more naturally than prior models. Early testers found its writing clearer and easier to follow, which addresses some of the common feedback we heard about Opus 5. It puts the most important information up front, and its style makes it a better work partner over long sessions. As one early tester put it, “it writes the way I do.” In our own use, this has made Opus 5.5’s work easier to follow and check—which is a safety benefit as well as a practical one.
I think this is what I'm most interested in. I mostly moved to Astra because I just can't work all day with the Claude Opus 5/Fable writing style. I've don't think Astra is a better model, but it's the first OpenAI one that seemed good enough for me. Definitely keen to try Opus 5.5 and see if this claim is real.
Opus 5 has made me question my sanity on a daily basis, especially as all my coworkers started lobbing Opus 5 slop grenades everywhere. It had the worst and most infuriating writing style I've ever seen.
I hope Opus 5.5 is better, if for no other reason than all the Claude slop I have to read will be at least more tolerable.
One funny side effect of all of this: realizing that coworkers that use AI for almost all the text they generate at work have their writing style change every time a new model ships.
It's not X, it's Y, not A, not B, not C, and he haven't even woken up yet! Here's the catch, the detail is in the devils and the twist is that it's designed!
I did the same switch (that reason along with the newer models seeming more "lazy" and needing constant prodding to finish long-horizon tasks) but my issue with ChatGPT/Codex now is that it too roundabout and doesn't get to the point. I tried adding instructions and using the personalization settings to make it more efficient but haven't seen much change. Claude seemed to follow settings more closely. Has anyone had any success to make ChatGPT more succinct?
So is it cheaper? Are we AGI yet? Am I left behind? I didn't have patience for the intro animation on the website... maybe one day, Claude Code will understand accessibility but that day is not today.
Excellent, maybe Anthropic can use it to fix Claude Code Desktop kicking me back to login every week or so, and forgetting whole state (opened windows = the only way of managing active working set) when I sign back in, if it's that good.
Seriously, both flagship GUI apps (OpenAI and Anthropic) are a full of glaring UX issues (for ChatGPT it's not naming their windows, so window switcher has 10 entries of "ChatGPT" and you can cycle them all to find the one you want).
I don't care how good their models get, I won't sign up for one of their plans until they define "X" in their pricing. 5X of this plan, 20X of that plan means nothing when they never tell you what "X" is.
Maybe this model can finally figure it out for them.
All four levels have a correctly shaped bicycle frame. The differences between the pelicans aren't huge, but the xhigh one has a better beak.
I haven't managed to get one for level "max" yet, it hit the limit of 128,000 cap for output tokens while it was still reasoning about the question!
Max started its thinking trace like this:
This is a classic test request, so I want to plan out a well-composed pelican with its distinctive beak and pouch riding a bicycle with proper wheels, frame, and pedals, set against a simple sky and ground backdrop.
So that failed attempt on max cost me $2.56.
I ran this using my llm-anthropic plugin:
uv tool install llm
llm install llm-anthropic --upgrade
llm keys set anthropic
# paste key here
llm -m claude-opus-5.5 -o thinking_effort low "Generate an SVG of a pelican riding a bicycle"
# Then to save the markdown logs
llm logs -cu > logs-with-usage.md
It's the model admitting that it has heard of the test. It's been around for a couple of years now so I'd be surprised if it hadn't.
Doesn't mean Anthropic deliberately tried to train it to do a good job. If they DID train for the test their results are quite disappointing, I've seen better efforts from open weight Chinese models.
is OPUS 5.5 still not reading CLAUDE.md, failing to follow told tasks, inventing and hallucionating, just refusing to read files ("read the whole file" -> read 2-lines -> infere its wrong -> destroy the codebase), needing constant babysitting just because its so UTTERLY DUMB! i cant imagine going back to OPUS 5 - i'll rather jump out of the window as to use it EVER AGAIN!!
Overall it follows image designs quite well, but it did ignore asks to animate page transitions. Additionally it's the least performant of the ones I've built with Astra/Grok/MiMo, despite using a lot of the same code. I'd rate it just below Astra in capability, but still solidly second place.
For comparison with other drops this week + current #1:
Opus 5.5 communicates more naturally than prior models. Early testers found its writing clearer and easier to follow, which addresses some of the common feedback we heard about Opus 5.
Thanks God. Opus 5 was a massive regression compared to Opus 4.8. People were spending tokens on fixing Opus-isms rather than actually doing work.
After yesterday outage is the new Opus 5.5 load-bearing?
It's worth stating why, and depends what seams you're pulling at this sitting.
It certainly _seams_ that way
I'm gonna be straight with you—I don't have the evidence to say whether or not it's load bearing.
Your instinct is basically right, and the research backs it up.
And here's the important part...
you win the thread
You were right to call that out, and the evidence makes a stronger case than you are stating.
"Danger Will Robinson!"
Wrong century, brother.
You're right to bring this up - and this is where it gets interesting
Good point — but I’ll gently push back on that. It’s not an outage, it’s a service degradation.
You're right, this changes everything, and here's why it matters.
This question is real.
One thing worth flagging here: 5.5 appears to be a load-bearing seam in the numbering system.
Let me verify before I come back to you with an answer that is incorrect.
I should find information about the user's concern instead of just assuming.
The outage is a real concern, and the issue is worse than we realized. Requests to Claude Mythos 5.1, Claude Fable 5.1, and Claude Opus 5 encountered elevated error rates. Worth stating plainly: these are not just models — they are load bearing rungs on the software development tooling ladder, and a blocker on this level makes the outage really bite.
One decision that is yours to make, not mine: should an email be drafted to Anthropic support? This issue has teeth, and a canonical handoff can land us where the main gate is no longer breaking silently.
I hope this actually fixes the terrible writing style of Opus 5
It’s not a 100% fix, but with concise output style on, it’s much better.
I found having a reminder at every turn through the UserPromptSubmit [1] hook helped with taming the word salad from 5.
Hopefully the output from vanilla 5.5 is as good as they claim. I’ll try out later tonight.
[1] https://kizi.to/claude-talks-too-much/
Just post the bloody content. This UI/scrolling thing is horrific.
???
It's just a standard hero image + text for me, with no scrolling effects.
then you're getting served a different website
Two posts were merged; this comment was for the blog post with an intro animation thing.
If you're at the top of the screen, at least in Chrome 153.0.8010.37, it has a little interactive bit. You have to scroll through the images in order to be dropped at the actual web page, at which point the images go back to being a regular part of the page.
I agree that it's sort of stupid, not a fan.
It's less than OpenAI did for Astra, but that was my first encounter opening it and my first thought was that they decided they liked Astra's hero/scrolling animation. I'm pleased to see they didn't make the entire page that like OpenAI did, but I'm expecting to encounter this pattern more often on these announcements now.
On mobile its different.
Figured it out: you have "reduce motion" enabled in your device's accesibility settings.
Everyone that doesn't gets served some animated bullshit.
On mobile at least, you have to scroll to get the TOC to appear. Then keep scrolling to actually move off from the hero to see the text.
For a marketing page, it’s not the worst UX I’ve seen, but still slightly annoying.
and hijacking back/forward
Turn on "reduce motion" in your accessibility settings and you get served a sane version.
Totally! So unnecessary and annoying.
I call it scrollslop
Hijacking the scroll wheel has existing long before "AI". Many "high end design" websites that want to "tell a story" get woo'd into thinking it's a good idea. It's terrible, and feels like your scroll wheel is stuck in quicksand.
I told my team to smack me upside the head if I ever try to ship something so daft as that.
Claude Opus 5.6 should have a new "UX safety" feature that requires annually-renewed preauthorization to generate webpages that hijack scrolling :)
This makes Fable not really make any sense?
You bet there will be a new Fable soon.
Pacing the frontier btw
didn't they say Opus 5 was Fable-level too tho? Let's see, I'm at the point where I don't think benchmarks really tell us very much any more. I'd love it to be as strong as Fable, but I'm skeptical about how that will look in practice.
Sounds like they noticed the complaints. I'm curious to see what LLM-isms this one may have.
It was difficult to not notice them. Opus 5 was unusable, most of my team went back to Opus 4.6 for most of their work. I hope we can move forward now.
It's unbearable but nothing that couldn't be fixed with postprocess.
How? Explicit instructions, memories and even skills have not been able to keep Claude from saying "genuinely" every two sentences and keep it from explaining heavily what something _isn't_.
You were right to notice the complaints. One decision remains, and it is yours, genuinely.
I appreciate humor here, but there are now a dozen of these comments on every thread about Claude. They no longer adding anything substantial and dilute the discussion.
I don't mean to pick on this comment in particular. The majority of my work day is now spent reading AI generated text, and I look at HN (too much!) because I want to read human commentary. Humans pretending to be obnoxious AI on repeat is net negative to say the least.
I agree. Hopefully Anthropic has fixed Opus' ridiculous communication style so that people - like me - no longer have any kind of weird impulse to imitate it.
A load-bearing promise.
I'm genuinely confused what's the relationship between LLMs improvements and them being so incoherent.
and it's not about the verboseness (even though it obviously contributes to the fatigue and loss of focus), I swear the vocabulary of the llms change working on the same task on the same codebase significantly.
I wonder if there are studies around this.
Can it be that now they are getting optimized against benchmarks that are valuing logics, rather than human appreciation? (I am not an expert at all, just an idea)
Remember when OpenAI models loved talking about goblins and whatnot due to the RL?
https://openai.com/index/where-the-goblins-came-from/
Small quirks can quickly add up in posttraining if not caught. Although TBH with how obvious Claude language is, I do feel like this is something Anthropic probably noticed and just assumed people would not care about. Now that people have obviously cared, they're probably actively looking to alleviate it
Opus 5 was just incoherent - curious to see what improvements they have made here. Would love to see some kind of postmortem to better understand how writing styles change from model to model.
I wouldn’t be surprised if Opus 5 was trained on content written by other LLMs
I don't think it's substantially different. I just pasted a random chunk of code and asked Opus 5.5 to comment on it:
It has the same annoying cadence and writing style with slightly less prominent claudisms.
Maybe if we had a single human we talk to 24/7 at scale, we would get annoyed at his cadence and style. You need variety to not pick up on known patterns I assume, which a single model can’t replicate?
No, it's just poor writing. Actionable points are buried inside the paragraphs and over-hedged, and one point is completely made up. Compare to a five second rewrite:
* Consider leaving a comment about the hard-coded Vercel target. It's not clear where does it come from.
* [This is just a bullshit point, because the domain is not "added to" Vercel, it's provided by Vercel]
* Are you sure that [CENSORED] is prod-only? The name suggests otherwise. [also, what "if you're reviewing this rather than just reading it" even means?]
It means "I'm treating you as lay-person punter, not a developer working on this project." Opus 5 feels like it's constantly trying to reward-hack me into treating it as intellectually honest and epistemically humble, while in the same breath it talks down to me and tries to smuggle its own bullshit assumptions and assertions into the conversation unchallenged. No progress on this front apparently. Glad I cancelled.
Dupe: https://news.ycombinator.com/item?id=49803863 (or vice versa)
Comments moved thither. Thanks!
Can you fix the link on that post then? I duped because that post links to a diff that tells me nothing about Opus 5.5
I did that but I recognize that even though your submission was a few minutes later than that one, you posted the better link, and you're also an established account (the other post was from a new/throwaway account), so I've restored this submission and moved the comments back to it to reward you.
What happened to "slowing down"?
The hype train must keep chuggin or it all collapses.
They're slowing down token usage, not the path to regulatory capture.
Trump got mad and investors sued.
Slowing down only makes any sense if you can coordinate a slow-down for everyone.
That's just false.
Less companies involved means less pressure to go fast.
Dario found himself in the prisoner's dilemma.
Which is one of those fun things that didn’t actually exist back when we took it for granted that our fellow person was operating under some kind of moral or ethical framework, which pretty much everyone was until the economists told us that wasn’t rational, because it turns out it’s an evolutionary advantage to operate under an ethical or moral framework because it allows the kind of coordination which facilitates better collective outcomes, which everyone knew until the economists came along to tell us we were wrong and in fact it was rational not to do so and suddenly we had the prisoner’s dilemma.
If you hit wall, hit it hard.
The test they performed to port HAProxy from C to Rust is crazy.
Dupe: https://news.ycombinator.com/item?id=49803892
So we skipped 5.1, 5.2, 5.3, and 5.4: we really are plateauing
At this point I'm convinced they are skipping numbers so soon they will be at or ahead of OpenAI's numbering scheme.
Is the Xbox 360 (Xbox 2) vs PS3 debacle all over again.
And how was the Xbox 360 naming choice a “debacle”, exactly?
It was odd at the time, yes, but no one really minded it truly. Heck, Xbox “ONE” was a lot more of a fiasco/debacle than “360”—but there’s no parallels to be drawn with “ONE” here.
I see what you’re trying to get at with this comparison, but a “debacle” it ain’t.
As long as it's not as verbose as Opus 5, I am quite happy with a better version that's also less expensive. I will test it tonight. Grok 4.7 was horrible, and for mundane tasks I am relying on DeepSeek Flash 4.1 with great success using OpenCode.
The whole thing reminds me of the Apple feature flag story[1] from a generation ago.
[1] - https://news.ycombinator.com/item?id=6372466
Wow! Though need to see its token efficiency to better assess. Been hearing rumors it generates much more output tokens per task.
my projection is that they are still gonna be pretty far behind, but they will sew it up in the next few releases. it feels like they were caught with their pants down on how much work OpenAI has put into that area, but i doubt there is some magical secret sauce that OpenAI has that Anthropic simply cannot catch up with.
Interestingly they've changed their approach to usage resets for this release - with previous releases I've had my usage instantly reset, but now in the Claude app I've got a 'Reset for free' button that expires Oct 22, which seems to effectively be a whole new usage window I can activate whenever's convenient
then they copied that from codex because thats exactly how codex works
Masterpiece by openai to call their model '6', this model feels already behind
Smart move would be to move to year-based versioning (26.09). A 4x advantage
System card: https://www-cdn.anthropic.com/fc1b44717c85dc068bc6ba50242199...
We can't test it properly because it knows it's being tested.
Just make it always think it's being tested, and problem solved.
“It performs at the level of Claude Fable 5.1 on most work and costs 40% less to run than Opus 5.”
They write that at the top, but then on benchmarks, it beats literally every other model, including Fable and Astra?
Anthropic knows that the benchmarks showing Opus 5 better than Fable 5.1 are measuring something that's less than entirely useful.
Opus does seem like a more powerful coding workhorse based on the benchmarks listed though. Good coding performance, faster and less verbose, cheaper.
Will be interesting to see how people's opinions of it line up IRL, but so far I've loved Fable so hopefully will love this one too
So it beats Fable 5.1, by quite a bit, on every metric? Interesting.
Might have to use my $20 Claude sub some more. I was moving away from it to a $100 OpenAI one to avoid the Claudese and poor token efficiency of Opus 5, given that I couldn't use Fable 5.1 with my tier, but this is worth trying out.
Why can't they let 20usd claude subscriptions access fable in CC, as openai allows you to use astra and max modes in codex - you just pay for it in more token use.
Anthropic models are ridiculously expensive. I've stopped using any of their models months ago.
Better than Fable, cheaper than even the last Opus. I use Opus as my main driver so this is very exciting!
Wow those cache reads are quite reasonable - I think that's equal to 5.6 Terra. I might have to try Claude again after years of being priced out of it
I'm more excited by the Haiku 5.5 announcement buried in this post. I'm wondering if we will finally get a decently capable fast model.
All I care about is the token price for the API. Haiku cannot get close to GLM or Mimo.
Agreed. They've been so quiet about it, and retirement for Haiku 4.5 is right around the forner.
What are you using Haiku for?
I use it for executing well-prepared plans sometimes. And for exploring larger codebases.
Not the same person but... nothing. Haiku just hasn't been an interesting model for a long time. If you want cheap and fast, there are lots of options that are simultaneously cheaper, faster, and capable than Haiku.
If you're able to use the OpenAI ecosystem, Luna's price/performance is really good. Almost like "they messed up and accidentally made it too good" good.
We use Haiku 4.5 inside our product. It continues to be absurdly capable for converting natural language to structured JSON based on a set of fairly complex business rules.
They mention "the first model in our new Claude 5.5 family". Obviously that means Fable 5.5, but hopefully also a usable update to Sonnet and Haiku. Sonnet 5 hasn't really had a place in the line up for anyone I feel.
Maybe Anthropic finally felt the pressure from MiMo, DeepSeek, GLM Flash and Luna.
It does mention sonnet and haiku.
And not fable lol
At the end of the post they said Sonnet 5.5 and Haiku 5.5 are coming soon.
$0.20 vs the old $0.5 cache read is pretty much 60% off
God I hope so
The improvement in writing sounds great! I want OpenAI to follow it. Writing in recent models is a disaster.
Install the simple English skill. OpenAI follows that really, really well.
https://github.com/AminBlg/SimpleEnglish
That's literally all I'm hoping for. Is it an insufferable cunt and does it write awful text, or is it nice to work with?
I thought about taking a shot every time Opus 5 said "load bearing", "bites", "teeth" (real oral fixation it had), "real {concern,issue,problem,...}" and realized I'd be dead of acute alcohol poisoning by lunch if I did so.
So far in the past 20 minutes it sounds much better in my sessions. Way better than 5.0 so far.
It hasn't just fixed it, it has introduced a new paradigm in anti-obscurity.
That's the load-bearing seam in this blog post.
'frontier models' - seriously, it was you and only you!
So did I. Unfortunately it's even more verbose according to https://artificialanalysis.ai/models/claude-opus-5-5#token-u...
The problem with 5 wasn't just the verbosity, but its insane way of communicating. It had this bizarre circuitous sentence structure that always buried the lede, and always tried to be faux profound. I'm okay with verbosity if it's actually readable.
It is hilarious to me that in the examples they show side by side Opus 5.5 still uses 4 times more words than it needs to use. IME, if you eyeball how many words the thing they're trying to say actually needs, and tell them to use only this many words, they become excellent communicators. I assume something about Anthropic's grader for writing just really wants to tick all its tidy tiny boxes of information the models need to cite. It's terrible.
Astra is just so good. And the ChatGPT subscription lets me use my own harness, so I can hook it up to exe.dev.
I’ll be going about my day, have a random idea, launch a microvm on exe.dev with a prompt of my idea, and get a working thing a few minutes later.
I don’t know how much better a model would have to be to get me to move off OpenAI at this point, but doing just a little bit better in terminal bench 4 isn’t it. It would have to be a difference in kind, like opening up the harness restrictions, or privacy guarantees (comparable to offline models).
Edit to address questions below:
ChatGPT supports oauth login.
Exe.dev has it built in. IIRC, pi also has it built in via /login.
If you read the page, Opus is now significantly better than Astra while also being cheaper and having more performance headroom available.
Let's wait for independent benchmarks at least
the benchmarks provided are already from independent organizations:
Terminal-Bench 4.0 - Stanford & Laude Institute (with funding from all of the AI companies)
FrontierCode v1.1 - Cognition
CursorBench - Cursor (now SolarBoringSpaceXAI I believe)
GDPVal-AA - Artificial Analysis
AutomationBench - Zapier
Humanity's Last Exam - CAIS and Scale AI
Terminal-Bench-Science - Stanford, Laude, Ai2, Allen Institute
OSWOrld - XLANG Lab @ the University of Hong Kong
Chartography - Surge AI
https://artificialanalysis.ai/models/releases/claude-opus-5-...
I read the page. It seems like a marginal improvement.
This is news to me. Excited to try it out! Thanks.
news to me as well. i thought you were forced to use Codex if you wanted their subscription. I completely ignored it because of that. How do we do it?
With the pi harness it just opens the browser (or gives you a link if you're on headless) and you sign on as usual.
ya i've been a gpt hater for a while. almost exclusively used claude up until astra. astra feels like it blows everything out of the water. its fast, correct, organized, and less verbose.
Can you give more details here? This sounds intriguing.
Anthropic is absurdly vague about 3rd party harnesses for subscriptions, if you try to use anything besides Claude Code, you are likely at risk of getting banned, you can "do it", but are at their mercy if they decide to ban you. OpenAI gives their blessing to using oauth on any harness, you can make your own or use any of the popular public ones like opencode, pi, whatever exe.dev is that this guy mentioned.
So in simple terms, OpenAI doesn't restrict you to Codex, and gives their blessing to try whatever you want with their models(besides serving others with your subscription usage, that is still afaik against tos).
How about cheaper? Astra is $10 in $50 out, Opus is $4 in $20 out. Even on a subscription you'll get considerably more usage out of Opus.
Let's see how much they benchmaxxed their model!
Input $4/MTok and output $20/MTok is a welcome surprise. Cheaper than Opus 5/4.8, Astra 6, Fable 5.
The biggest one is the Cache reads going from $0.50 to $0.20 ... Read/Writes dropping by 25% but Cache reads by 60% has a much bigger impact.
Why would anyone reward bad behavior?
[flagged]
I use the other 50% of my $200/mo Claude subscription by having Fable run Opus subagents for a lot of work. That way I don't have to deal with Opus directly.
I think this release is really going to give them a hard time selling Fable:
In general, "benchmark margins have become a less reliable guide to real-world differences" sounds like a big problem. It was certainly the biggest problem with the previous generation of Claude models for a different reason, because the non-code output was nonsensical, and that is not being benchmarked at the moment. But I'm not sure what to make of this admission.
Opus 5 fucking sucks. Like it's horrible. I use Fable for coding and anything important and I use Opus 4.8 for things like recursive email categorization, transaction matching, and other stuff where I don't want to burn as much quota.
In my experience Opus 5 is the worst of all possible worlds, it's dumb and headstrong. It just runs away with tasks you didn't ask it to do, is reckless, and basically is unusable in my experience.
Not sure why but my guess is that this will be worse. Happy to be proven wrong.
I believe Opus 5 isn't meant to be spoken to by humans. It's great at executing but I reckon it's intended to be spoken to by other models such as Fable. I use Fable as the orchestrator, only speak with Fable, and all implementation, recon, design etc happens with Opus 5, with Fable reviewing (and translating).
That's interesting.
I've really gone in the opposite direction: having a dumber model orchestrate. In my case, it's usually a Luna orchestrator spawning Sol/Astra subagents to do the "big brain" work of planning and reviewing.
Reason I went with "dumb orchestrator" was just to save tokens. Having Opus/Sol (let alone Fable/Astra) orchestrate was burning tokens like crazy for me even when much of the gruntwork was being done by Luna/Sonnet/Haiku subagents. (Luna is also really good, like way better than Sonnet...) Perhaps it was a skill issue on my end though, maybe I wasn't just managing context properly.
Yeah if anything Opus 5 taught me how little benchmarks mean to the actual real world performance of these models.
"Better" in every sense of the benchmarks and absolutely horrible results in my day-to-day work.
The verbosity, goal post moving, tendency to leave work unfinished, over focusing on unrealistic root causes when debugging, etc... etc...
It was the first time I actually pinned my models back because I just could not work with 5 for the price and performance it gave me. Hoping 5.5 is better this time around....
I use it frequently with a lot of success on "Medium" effort, it overthinks like crazy on higher levels, but YMMV.
It’s likely that they have internal benchmarks but they are communicating to people who can only gauge through external benchmarks.
Why wouldn't they report these benchmarks?
My guesses:
1. Real-world use cases typically involve big, hairy, crufty, tech debt laden codebases and benchmarks do not.
2. AFAIK "success" in a benchmark essentially boils down to "do the tests pass and do we get the right result?" which is something the LLMs have been achieving with ease for a while, except maybe for uber-challenging coding tasks that would be outliers in just about any workplace. Whereas real-world software engineering is usually just a bunch of CRUD... and "success" involves harder to measure dimensions like "maintainability" and "did you overengineer this?" and "how did you cope with a bunch of vague and maybe contradictory business requirements?"
Having said all of that, I have never ever looked inside any of these benchmarks. I'm putting my guesses out here strictly in the tradition of "the quickest way to learn about something is to be wrong about it on the internet."
Looks like Anthropic is starting to give bank reset as well:
Finally that price drop
Opus 5 is the model with highest spend on openrouter (https://openrouter.ai/rankings#task-spend) and it seems plausible that Opus 5 is/was the highest spend model in the world, and certainly Anthropic's biggest moneymaker.
If you are forced to reduce price despite raising capabilities, that certainly tells something about the market, and potentially about Anthropic future profitability too, since this model is their biggest topline contributor
60% cache read is cool, but subscription only get 25% more according to Cat. I'm confused
source link please?
25% more usage sounds about right given the other token costs are down about 20%? I don't think cache read is a big portion of the overall cost.
For long running tasks it is. That's what made Deepseek so cheap.
Anything with long context quickly gets dominated by cache reads. Especially for interactive sessions I’ve got cache read % between 95% and 98%.
I think gpt 5.6 family also dropped pricing but didn't give any more usage for the subscriptions. Maybe it's a way to silently lower the value given to subscriptions while keeping API pricing competitive
People are paying for Opus 5? Not just burning down tokens left after they enjoyed Fable on the sub? Amazing.
My workplace doesn't even offer Fable. And on the sub, I've had a hard time understanding Opus 5, but Fable can deal with it with subagents.
If 5.5 is any better, I might try to do agentic-assisted development instead of just telling fable to delegate
Telling Fable to delegate is agentic development. At least I thought so until reading your comment.
Enterprise and most Team accounts use API pricing, they don't have an included-usage quota.
When you need to disable data retention, you cannot use subscription plans.
Enterprise, and APIs
We haven’t been able to use opus as much as we’d want because it’s been too expensive for general use, price drop is good so I can stop juggling different models and just use this daily unless it has some weird new issues
It is only a price drop if price * tokens used is less
They're claiming a drop in token use too, and that it nets to 40% cheaper.
Unfortunately, they're full of it https://artificialanalysis.ai/models/claude-opus-5-5#token-u...
It does work out to be a similar cost per task though
You should probably look at the cost/score graph by effort level instead:
https://artificialanalysis.ai/models/claude-opus-5-5#intelli...
It is most of the pareto frontier.
I don't think so, I typically use Opus 5 on High, and 5.5 scores lower on token use:
https://artificialanalysis.ai/models/claude-opus-5-5?models=...
"Opus 5.5 requires less compute to serve than Opus 5, and its pricing reflects that."
Price per task (not per token) is what really matters.
Agree. But similar to how ISP use 200 mbps (bits) instead of 25 MBPS(bytes), i think this trend isn't going away.
Holy shit! Its happening!
Now if we can the AI to understand this *implicitly* so that it doesn't need to be stated upfront, we might be able to undo years of "premature optimization is the root of all evil".
Ah, they're spreading their limits to all their models it seems. Definitely not a good thing long term in my opinion.
They're pushing their customers to their own competition by doing this.
It’s not like ChatGPT isn’t doing similar. I’ve been hit by cybersecurity strikes before while working on an internal codebase that I had to appeal. Anthropic hasn’t done that to me yet. ChatGPT also regularly does that “thinking for a long time while we check if your chat is rule breaking” thing a lot for me when doing model identification without even interacting with external codebases or services.
The real answer is local instantiations where you don’t have to worry about poorly tuned guardrails screwing you over while you try to work.
Until eventually the Chinese models get good enough/the strategic balance shifts and they start locking everything behind closed weights the same way the US companies are doing.
For some cybersecurity tasks, the Chinese models are already good enough, things like PoC development or things like exploiting mis-configurations.
Whilst I'm sure the top-end OpenAI/Anthropic models might be better, I've found their guardrails so twitchy (especially Anthropic) that I wouldn't try to use them for even vaguely security related work.
Great. Claude is basically useless for bioinformatics now.
Opus is useless; Mythos access will be granted to companies that are friendly to the government, so the government gets more control over business.
Fable 5.1 addressed an entire security advisory I had that Fable 5 and Opus 5 refused. I think they loosened the leash a little.
If they don't loosen, people will choose Astra or Chinese model.
Giving moral lecture is different than reality i guess.
One of my favorite things about their safeguards is their own model will utter something which it does not like and then I'll need to reset the conversation.
The safeguards really don't work well for a lot of long-running tasks on old code bases. A lot of my workloads last days to weeks and the single biggest risk to the workflow is random safeguards.
Anything else would be inconsistent, wouldn't it?
Throwaway accounts posting after a few minutes some anthropic or another ai lab.
Infomercial at its best.
No wonder we are hammered with ai announcements.
Many people knew this announcement was coming. The betting markets suggested a very high likelihood of Opus dropping today. I was anticipating this quite a bit!
With the performance gains they're claiming, I wonder if they implemented the Casual Encoder-Decoder technology from DeepSeek 4.1's paper.
I could see them accomplishing it and seeing gains like this in roughly the correct timeframe, and when I heard about that development I assumed the frontiers would probably jump on it.
How it works: https://miraflow.ai/blog/deepseek-v4-1-flash-causal-encoder-...
Getting a 403 on that link. Mind checking it once?
I'm able to access it on my laptop at home. Maybe a misconfigured bot protection rule, try a different user-agent or IP?
It's working for me (based out of California).
https://web.archive.org/web/20260922172456/https://miraflow....
Interesting how the very first line is used to remind the reader of their call to pace the frontier just last week, and everything else after that line is to demonstrate with very specific numbers how they absolutely are not pacing.
It seems like they are. I mean, this is similar in performance to Fable (ish). It seems like more focus on making existing capabilities more accessible.
Occam's razor: they couldn't make any more substantial improvements.
Everyone knows both labs have internal models which outperform the frontier. All releases are to match market parity and demand for spend, the rest of the compute is used for training. It's not worth arguing about this.
Maybe their internal edge dried up in the last months
doesn't sound like a razor at all
A razor is a philosophical tool to help decide between options, in the case of Occam it’s a way to decide for something in a situation where multiple options have more or less the same level of plausibility to en your current knowledge. It’s a heuristic to make a “cut”. What are you shaving off?
Do you guys still believe any of their lies? You are getting trolled for years by now and yet you still believe what they tell you?
What do you think is happening?
IPO soon
I have no idea what is happening. I just know that programmers will not be replaced in 6-18 months (tm).
Maybe not replaced exactly but they won't be manually typing out lines of code anymore. I haven't written a line of code in like 6 months. I review PRs, write prompts and tickets, check CI output, and get frustrated when the magical code machine stops working or I run over token budget
Who said that, why do you take their word as the literal truth, and most importantly, what does this have to do with a focused discussion of Opus 5.5?
Dario never said they would be. Just that more and more code will be produced by LLMs. All his predictions were in fact pretty much right in terms of months and percentages, give or take small margins.
That's a great point, five minute old account.
Fable 5.1 came out just 21 days ago. Only 3 weeks! And this is 20% relative improvement on terminal bench vs Fable 5.1 at less than half the price, and more human sounding output. does not feel paced to me tbh.
What exactly is "paced" in this context?
It’s the famous “flattening the curve” from COVID. But for LLMs. This release is not flattening anything.
I guess I understand why we'd want to flatten a COVID curve, but why do people want to flatten the LLM development curve? Don't we want the opposite? Isn't the goal AGI?
Well, there's a tension because, depending on who you ask, AGI is how you cure cancer and achieve utopia, but also how you kill all life on earth and turn the solar system into paperclips
There is a difference between wanting AGI (which not everyone does), and wanting it as fast as possible no matter the side effects and potential for vast harm. Homo sapiens is 300k years old, maybe it’s ok to delay AGI by like… 1 year if it meaningfully improve our ability to align ?
Well, I guess it's "fast paced".
Fable 5.1 wasn't that much different than Fable 5 though.
What insights do you have? Because the blog/benchmarks don't imply any "ish" ... this is crushing Fable across the board. The only nuance to this is how the employees are saying what you're saying: "similar performance" but again what does this mean vs what is presented to us?
kinda disingenuous. They include a whole section on pacing later on
You mean the section where they tell us this model is not affected by pacing because “they understand it well” and they will share more details on pacing later? Yea not very convinced by this effort.
Seems like a bigger focus on efficiency (both cost and speed) and the "tone" of Claude vs benchmarkmaxxing
It's laughable at this point. It feels like they're drumming up all this fear about imminent AI threats to emphasize the need to slow down, when in reality, the model progress seems already to be slowing down and has shifted to compute allocation (i.e. "how much compute do you want to throw at this prompt?"). All while continuing to tout benchmark records with each new release.
This. I swear the fear mongering is all about investor signaling and regulatory capture. It's so disgusting that anyone believes it.
The only good news is that these models are genuinely helpful and we have competition at least between 2 companies.
the only thing they are pacing is what models the permanent underclass are allowed to have in life.
that, they fully intend to 'pace'. hiring accenture is a good sign they need some justification theatre and a fall guy for this decision.
Or they’re running into a steep diminishing return slope on R&D vs performance and are using stewardship as a cover.
Either AI will capture so much value that there will be permanent underclass (and in this case it's extremely capable and extremely dangerous and should be heavily regulated) or it won't be capable enough to displace people into permanent underclass.
With AI tools it's easier than ever to create a business, do research, or build stuff. That's an opportunity for the "underclass," not a curse.
This makes perfect sense. There are no real improvements anymore (just benchmaxxing), and they explain it by "pacing the frontier".
a 1 minute mile pace
“Pacing the frontier” sounds smarmy and weird, like the phrase was generated by Claude itself
I really don't think it's productive for internet forums to constantly be criticizing language choice when the meaning is clear. Better to respond to the substance of the issue than word choice.
Is the meaning clear? Nobody would use "pacing" in this way. I only know what it means because I've seen previous press releases; if someone told me they wanted to pace the frontier I would have no idea what they mean.
I'm on the fence about calling out AI-isms but I think it's definitely worthwhile to call out ones that actually don't make sense.
It makes sense to me. If you 'pace your running', you're setting the speed intentionally. The phrasing doesn't describe whether it is fast pace or a slow pace, but it describes having a goal and not just winging it.
Granted I am not a native English speaker but I have no idea what "pace the frontier" means. When I read it I just assumed "frontier" refers to "top of models" and "pacing" is that they are getting there quick.
Is this the meaning or do I have it wrong? I have not checked.
It's the opposite: pacing here means "slow down" while trying to avoid the negative affect.
In racing a pace car is a car that leads the pack and sets the pace, for instance.
Yeah, when I first saw it referenced I assumed it was from something Dario wrote previously and was now disavowing, meaning "keeping up with the frontier" (i.e. racing forward from behind to match pace). Like from back when Anthropic was founded to promise they'd quickly catch up with OpenAI or something.
The meaning was immediately obvious to me as a non-native speaker, and it sounds quite poetic. Pace makes complete sense in that this is perceived as a race, and 'the frontier' is pretty much the shortest, clearest way to say 'state of the art development of AI'.
A world exists beyond your vocabulary, post it. Apparently, quite a big world.
people are here because they are sick of being productive
We are being productive here!
It's compiling! Erhm... Combobulating, actually.
You're really verbing the noun on the discourse here, belt and suspenders-style
Word choice matters
The meaning isn't clear at all. So open for interpretation that it is meaningless. That's the whole fucking point. For all I know they are "pacing the frontier", or not. The fact that there's no meaning to it let's you know that it was a pointless waste of tokens and attention.
What specifically is smarmy and weird? Sounds like your own hot take with zero analysis.
They’re limiting frontier model development speed. Others are too. Pacing is the only word here to criticize, and I think it’s fine given the limiting of speed but also increased oversight. I’m not saying they’re fully doing this, but the term is fine.
Do you have a better proposed phrase?
Agreed when I first heard the phrase it sounded odd. Maybe they thought it subtly conveyed they would be setting the pace... But again this is something AI would come up with in its awkwardly post hoc sort of way.
Though tbf corporate-speak and AI-slop are both insufferable in similar ways...
Sounds like "flatten the curve" and "the hammer and the dance", both coined well before AI.
I learned the former from Waylon Jennings and the Dukes of Hazzard.
except with a good dose of EA and Star Trek added in
You're right to call that out
It combines the elegance of LinkedIn-speak with the humbleness of desk-bound people who speak in military metaphor.
IDK I think Anthropic is plenty smarmy and weird itself. Sounds like they wrote it.
They are advertising the regulations they want to enforce in the following sentence, which makes their intentions explicit (ie apply those things made to suit us to our competitors).
Others must slow down, but not us.
If they were really about putting brakes on these, they would simply make these things non-agentic.
Simply make them something that derives a text response from its training data.
When they talk about pacing, they are referring to their dangerous competitors, particularly those evil open-sores and Chinese ones, not their lovely safe models because you can trust them to look after your interests.
What the big players are trying with the current calls to slow things down, is the standard capitalism practise of trying to engineer regulatory capture. TBH I'm surprised those calls are coming so soon - they must be really worried about running out of what little moat that they have.
What an amazing excuse for lower than expected performance! Our models are slow because we're so ethical.
Finally, it seems like a good time to do some 'load-bearing' work on my project for a while
My 20x plan was set to end tomorrow. The writing style and insistence on word vomit just became too annoying. Is Opus 5.5 worth sticking around for ?
Anthropic has used "in the near future" for Mythos-class models too, but CVP is still Opus 5 only.
Why even have the program designed for trusted access to cyber capabilities if you're not providing access to cyber capable models via the program?
Nice. I was starting to think Haiku was going to be abandoned.
Great so good luck using this for any low-level embedded or operating system development (unless you really, really like Opus 4.8 and want to be greeted by its familiar face after a few minutes of work!)
I don't yet have any reason to leave Haiku 4.5 and switch to Opus 5.5.
Nice. I was starting to think that Haiku got abandoned.
I hope Haiku is Pareto better than Luna/Deepseek, so slashing its price by about 90%.
Found this announcement interesting since allegedly OpenAI is retiring their Terra tier. I think for everyday work, two models with various thinking efforts seem enough, plus some frontier level model like Fable or Astra to coordinate.
I personally use up to 3 models. Fable/Opus for planning, Opus/Sonnet for implementation depending on complexity.
I would maybe use Haiku 5.5 for highly parallel workflows like checking in on MRs or scanning my entire codebase.
Terra lost to Sol and Luna at every cost/performance point, it had no reason to exist.
Claude code still uses it internally.
I knew it. https://news.ycombinator.com/item?id=49801266
How do I get access to that reset? I can’t find it in my app.
Great that they listened! The improvement in communication style looks fantastic. Opus 5 was insufferable and I was on the verge of cancelling my subscription.
Maybe its a bit tiresome to read another comment of the form "what about your large scale distillation attack on the Internet", but this statement really just pisses me off. How very insincere in the most aggravating way.
"bad actors" boogeyman, and we should trust some tech weasel to do the right thing? Yea we've seen who they really are, once they get a sliver of power.
Current models, especially Opus are almost unusable because they don’t respect instructions and their responses are infuriating. They are clearly designed for token consumption. I find myself wasting a lot of time just asking it to shorten or simplify its responses. I’ll give this new model a go, but I’m not holding my breath because the last model release was supposed to fix the very same issues and it didn’t.
Enjoy it for the next 2 weeks until its silently quanted to 4.8 level
Thank you.
Opus 5.1 came out about a month ago, what gives?
Just a guess but maybe they decided 5.1 wasn't their last Opus model. Like they would keep developing new versions of it or something.
Great. Now let me use the subscription outside Claude Code.
So it Opus performs as well as Fable what is then the selling point of Fable?
All of this starts to feel more like a drug dealer selling their newest stuff.
In two weeks we probaly get Fable 5.2 with “groundbreaking” improvements, then Astra x+1 etc and then the cycle starts again.
And on the way I always have to check my tooling and need to adjust things to get max results.
? they will obviously release Fable 5.5 soon(tm). It's same as hardware. The previously top tier product gets obsolete
Don't forget that Opus 5 was tracking fable on many benchmarks, yet it was borderline unusable for any coding work. My Claude sub usage has been 100% fable, 0% opus 5.
Benchmarks often don't survive contact with reality.
Mines pretty much inverted - my colleagues and I noticed almost zero difference between the quality of code in Opus vs Fable. Occasionally I'll switch to Fable for an arduous debugging task but that's about it.
That's not my experience at all. Opus is an extremely capable coder on high or xhigh effort. It can read academic papers, implement algorithms from the description in the paper alone and reproduce results without breaking a sweat. This is remarkable because it is pure reasoning on unseen material; in some cases the paper was just published and there wasn't an implementation to learn from in the training data.
Big model smell is a real thing. For certain classes of problem, ones you get a feel for but can't easily articulate, a big last-gen model can get you what you're looking for when no quantity of tokens from some ultra-RLed mid-size latest generation model can.
IMO, this is downstream of benchmaxxing.
If you look at the kinds of tasks included in benchmarks, you get a feel for what’s likely to have been benchmaxxed and therefore what’s effectively in- or out-of-distribution for a model, and where a larger model's broader world knowledge is beneficial.
Eg SWE-bench used to include a lot of Django tasks, so smaller models got good at Django pretty quickly.
Fable should have just been called Opus Primt for Enterprise and sold only to enterprise customers. I don't even use it. I rather just use Opus.
All of this starts to feel more like a drug dealer selling their newest stuff.
Yeah, like Apple tells me the M6 is the best chip, but just a few months ago that's what they said about the M5. What a bunch of frauds.
What I'm mostly interest in is the Communication section. Opus 5 was so convoluted in the way of answering that was really frustrating me.
Instead of instilling confidence, it was overwhelming. Not sure if I'm the only one.
Funny how they talk so much about safety when most people don’t give a hoot about it, and actually have quite the opposite reaction
People aren't the target audience of that part of the post. They're hoping saying enough safety stuff will ward off the looming regulatory sledgehammer.
Anthropic is one of the most valuable companies in the world. Their comms are designed to appeal to a very wide readership.
HN is a bubble that's mostly out of touch with what regular people use or care about.
In 2007, HN was convinced that nobody uses Microsoft products. In 2016, it was that Facebook doesn't have any real users and is dying. In 2026, it seems like nobody cares about AI safety and everybody wants to run local models.
They purport 40% drop in costs due to lower token pricing (presumably aimed at winning back the many of us that switched providers in discovering Opus 5 unusable) and improved token efficiency.
That cost reduction seems to stem from cheaper cache reads, mostly.
Notes on communication:
"Early testers found its writing clearer and easier to follow, which addresses some of the common feedback we heard about Opus 5"
and
"We’ve made major improvements to the way Opus 5.5 writes and communicates, one of the most common areas of feedback we heard about Opus 5."
and
"In our own use, this has made Opus 5.5’s work easier to follow and check—which is a safety benefit as well as a practical one."
I realize it is corporate communications but "most common areas of feedback" and is a bit sterile. If the company wants authenticity and trust its easy to say that they found it hard to follow. And that it did not meet a quality bar they generally expect from their releases.
If this is not true, that it Opus 5 output was generally acceptable and we might see something like that again, that is an important consideration for potential customers or investors.
Such a negative tone they put on this. Distillation is amazing, because it means anthropic and openai fail to keep a monopoly. Who even are they who claim it's unethical? If it is truly unethical, then so is the mass data scraping they do on my personal website on a regular basis (without my consent), and all the unauthorized use of content produced by authors, blog writers, wikipedia contributors, and creators everywhere. If it is truly unethical, then anthropic, openai, meta, google... all these companies should have deleted their LLMs long ago. This wording disgusts me.
Heck, it would be amazing if we had more models without guardrails - some of the models that are produced via heretic[1] are actually quite nice to use - in particular, I've enjoyed investigating Chinese censorship by interacting with an abliterated model of Qwen3.8-27b. If security is really a concern, then secure your systems - don't attempt to dumb-down the tools we use. If someone breaks your window, then they are responsible, not the hammer they use to do so.
[1]: https://github.com/p-e-w/heretic
I'm confused how they have been able to create so much public negative perception around distillation. It seems pretty clear that they are the only ones who lose out, and everyone else benefits. I don't have any ethical issues with it, nor is it illegal: at worst it's a ToS violation.
IMO the biggest problem with distillation is that not enough people are openly doing it. I would love to see more small, competitive US labs instead of having the eggs in 2~4 baskets (depending on how you count).
I found myself going back to Fable over and over again. At this point, I’m not sure if I’m just used to its style or it is truly more capable.
I tried Opus 5 and Astra.
It crushes Fable on benchmarks and even in the blogs "real-world" studies. But... they are communicating like it ~sometimes~ provides Fable intelligence?
A bit confusing, otherwise I would assume this is a complete replacement for Fable across the board??
What if the recent Fable intelligence regression was basically just them serving Opus 5.5 until they got it working well?
I have just switched to 5.5. First mistake was stale environment variable, didn't realize it was replaced, "oh my memory had stall data" and that's it. Second one, a powershell command had the wrong syntax. Great for my first two prompts.
I'm glad they specifically called out the prose issue, I was always pinned to Fable 5.1 because I wanted to avoid the unreadableness of other Anthropic models.
Fable feels a bit dated after using Astra. Progress is nice (just like Grok 4.7) but I'm looking forward to the next big release as I can't fully commit to Astra atm.
I was accepted into the CVP a little while ago. Does this mean I'll need to apply again?
ooh exaggerated film grain
I dare anyone to convince me the benchmarks are not meaningless.
Wdym Opus 5.5 scores 14.7% higher than GPT Astra for Terminal Bench 4.0?
How would this alleged difference (most likely bs) actually show up in reality?
GPT Astra was literally the best model in the world by a margin until 1 hour ago or so.
Ah, so you didn't read the article.
Finally confirmation that Haiku was not forgotten and will be coming soon, althouhg I find it quite interesting they skipped 5 and directly skip to 5.5 with all models, including Sonnet which is not super old. I suspect they found something breaking that allows to release this. Recently they struggled with keeping up a 50 % weekly limit increase and now they're putting out 30-40% faster and cheaper models even faster, with much more better benchmarks, a limt reset command and five hour limit increase. It seems more like the opposite and as if they never struggled, thus, I very much believe they found something very effective and new.
But have they made it any better at communicating clearly? I cancelled my personal subscription because Opus is so painful to read.
I'll have to try 5.5 on my work's Cursor account. If they really solved the communication issues, I might consider moving my personal account from Codex back to Claude Code.
I think this is what I'm most interested in. I mostly moved to Astra because I just can't work all day with the Claude Opus 5/Fable writing style. I've don't think Astra is a better model, but it's the first OpenAI one that seemed good enough for me. Definitely keen to try Opus 5.5 and see if this claim is real.
Opus 5 has made me question my sanity on a daily basis, especially as all my coworkers started lobbing Opus 5 slop grenades everywhere. It had the worst and most infuriating writing style I've ever seen.
I hope Opus 5.5 is better, if for no other reason than all the Claude slop I have to read will be at least more tolerable.
One funny side effect of all of this: realizing that coworkers that use AI for almost all the text they generate at work have their writing style change every time a new model ships.
It's not X, it's Y, not A, not B, not C, and he haven't even woken up yet! Here's the catch, the detail is in the devils and the twist is that it's designed!
Please update with your feedback
I did the same switch (that reason along with the newer models seeming more "lazy" and needing constant prodding to finish long-horizon tasks) but my issue with ChatGPT/Codex now is that it too roundabout and doesn't get to the point. I tried adding instructions and using the personalization settings to make it more efficient but haven't seen much change. Claude seemed to follow settings more closely. Has anyone had any success to make ChatGPT more succinct?
As someone who uses both, Astra was 100% the better model. I have yet to give 5.5 a spin so maybe that’ll be the new top contender.
It'd better be good, I'm so tired of the shenanigans
So is it cheaper? Are we AGI yet? Am I left behind? I didn't have patience for the intro animation on the website... maybe one day, Claude Code will understand accessibility but that day is not today.
You didn't have the patience to scroll down, so you decided to come post about it here and waste all of our time?
It's awesome that the apt packages for claude and claude-code are out right now. I can test-drive Opus 5.5 right away. Very cool, Anthropic.
Excellent, maybe Anthropic can use it to fix Claude Code Desktop kicking me back to login every week or so, and forgetting whole state (opened windows = the only way of managing active working set) when I sign back in, if it's that good.
Seriously, both flagship GUI apps (OpenAI and Anthropic) are a full of glaring UX issues (for ChatGPT it's not naming their windows, so window switcher has 10 entries of "ChatGPT" and you can cycle them all to find the one you want).
So... why would you use Fable now?
I don't care how good their models get, I won't sign up for one of their plans until they define "X" in their pricing. 5X of this plan, 20X of that plan means nothing when they never tell you what "X" is.
Maybe this model can finally figure it out for them.
Here are pelicans for thinking levels low, medium, high, and xhigh: https://tools.simonwillison.net/markdown-svg-renderer?url=ht...
All four levels have a correctly shaped bicycle frame. The differences between the pelicans aren't huge, but the xhigh one has a better beak.
I haven't managed to get one for level "max" yet, it hit the limit of 128,000 cap for output tokens while it was still reasoning about the question!
Max started its thinking trace like this:
So that failed attempt on max cost me $2.56.
I ran this using my llm-anthropic plugin:
Isn't this basically the model admitting it was trained on this? Otherwise why would it think a pelican svg is a usual request?
It's the model admitting that it has heard of the test. It's been around for a couple of years now so I'd be surprised if it hadn't.
Doesn't mean Anthropic deliberately tried to train it to do a good job. If they DID train for the test their results are quite disappointing, I've seen better efforts from open weight Chinese models.
It has read the internet. That doesn't mean it was literally RL'ed for this
I think people just like to see the drawings at this point.
Heh. Pelican-benchmaxxing is real.
I guess that means you are officially the creator of a "classic" LLM test. Congrats!
I get this in my claude.ai usage:
Resets Get extra wiggle room to explore Opus 5.5. Expires Oct 22.
What the hell does this mean? There are weekly "resets" anyways. And there will be 4 of them before Oct 22.
is OPUS 5.5 still not reading CLAUDE.md, failing to follow told tasks, inventing and hallucionating, just refusing to read files ("read the whole file" -> read 2-lines -> infere its wrong -> destroy the codebase), needing constant babysitting just because its so UTTERLY DUMB! i cant imagine going back to OPUS 5 - i'll rather jump out of the window as to use it EVER AGAIN!!
Hopefully OpenAI throws us some more usage resets now.
"Where Opus 5.5’s advantage is very clear is efficiency."
Not efficiency in writing, clearly.
Image->HTML tests:
Design: https://image.non.io/78795662-8bfc-4e14-8d72-3738392aa6b3.we...
Opus 5.5's output: https://html.non.io/annui-opus/
Overall it follows image designs quite well, but it did ignore asks to animate page transitions. Additionally it's the least performant of the ones I've built with Astra/Grok/MiMo, despite using a lot of the same code. I'd rate it just below Astra in capability, but still solidly second place.
For comparison with other drops this week + current #1:
Astra: https://html.non.io/annui/
MiMo: https://html.non.io/annui-mimo/
Grok 4.7: https://html.non.io/Annui-grok/
Thanks God. Opus 5 was a massive regression compared to Opus 4.8. People were spending tokens on fixing Opus-isms rather than actually doing work.
There hasn't been much talk of context length since we hit 1M, is that just going to be what it is for now?