* From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
* On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.
* After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.
The "offer, then remove" aspect is a bit eyebrow-raising -- it feels like they are trying to get subscribers to switch to usage-based billing, which makes me wonder if we'll ever get it after that June 22nd window.
Still satisfied with my switch to codex/chatgpt. I couldn't imagine switching away from claude code when it first launch but with the drastically more generous usage on codex for the same subscription tier I just can't justify it.
My experience is that the GPT-family of models are very smart and figure out bugs, edge cases a bit better, but it produces code that is much less mergable – if you review the code, it introduces a lot more useless/inappropriate heavy abstractions and wrapper functions, compared to the Claude-family models which introduces the right amount of straightforward human-style code.
I can recognize so much of the GPT/Codex generated code long after it gets merged (not by me).
Additionally, the time spent on every agent turn on GPT 5.5 is much longer compared to Claude Opus 4.8, which means iterating on the code takes a lot more patience, and there's a lot more nitpicks to pick when actually using GPT 5.5 to do software engineering.
Feels like GPT-style models are more geared on doing one-shot software vibing (and handling the vibe coded mixture) compared to Claude's focus on actual software maintenance. I got a GPT Pro sub for free and wanted to cancel my Claude subscription so much, but I still keep reaching Claude models a lot more. Frustrating.
Have you tried iterating on style feedback in AGENTS.md? I've been reasonably successful using this to get it to output code in a terse, non-defensive style that matches my hand-written code.
I'm not sure if i do something differently but i have the exact opposite experience with these models. Claude always feels like it's generating way too overdesigned and hard to understand code with the vibe oriented feel while codex is cleaner and more "task at hand" and easier to work with.
This is my experience as well. I have defined a CLAUDE.md rule to ask codex to automatically code review, and I tell it that the reviewer is very picky and to only implement what it considers valuable feedback. I hope they don't converge over time, currently, in combination they works really well.
Codex IME is just smarter, I think it shows given both anecdotes but also how OpenAI has always been at the front of programming competitions and math problems.
But Claude models seem to be better at long term problems or more ambiguous problems.
I'm curious as to what the primary benefit here. Are there secret improvements in training? There hasn't been much in fundamental model architecture, I don't think. What about harnesses? I wonder what's pushing the AI. It seems like harnesses is the main thing pushing AI ever since CoT.
I find that OpenAI's agentic tools and models are better for building human-maintainable software. Meanwhile, Anthropic seems to be cosplaying Apple while missing out on all the exceptional engineering required to create something that polished. Their admission of predominately using Claude with little human oversight and their stealth mode is an indictment of a poor engineering culture, from what I can surmise.
Serious question: what is the secret to getting Codex to write decent code? I am on Windows. Maybe that is the issue, but I can't seem to get Codex to function anywhere near the level that I was previously able to get with even Claude Sonnet. Does Codex just not work well with Windows yet?
I've had the exact opposite experience. For various reasons, I've had to move from Claude to Codex and the rate at which it burns tokens for the same output I would get from Claude is ridiculous. I'm probably burning tokens at a rate that is at least twice as much as I was when using Opus 4.5 for coding tasks and still finding that just manually coding is easier than trying to get Codex to write functional code.
Agreed. I think the Chinese labs are proving that OpenAI and Anthropic don't have a moat in almost every aspect, especially pricing. I also think people are getting annoyed with the constant lift and shift. I've seen more folks drop Claude Code and Codex, specifically, because of the lock-in it provides the providers. I'm curious to see how people standardize on tooling adjacent and if Anthropic, Google or OAI move to block utilization akin to the games Anthropic has been playing as of late.
I think the end game is routed model usage and SLMs. I think Apple is going to prove this in the consumer space pretty handily and I'm curious how the Android ecosystem responds since the hardware is considerably lacking in model performance. I think Apple has a huge opportunity here, as much as I don't like their current ecosystem of walled garden. They did position themselves very well with ARM and custom chips for their hardware. Hopefully the broader ecosystem of ARM and Linux are able to make some headway and we see a more formalized, and broadly accepted, architecture to capitalize on.
My bet is they'll keep subsidizing for a considerable period of time, at least 1-2 decades more.
Most AI companies are just testing the waters with paid tiers right now, their greatest fear with increased pricing is folks reverting back to wikipedia, stack-overflow and other public domain organic activity buzzing back to life; that will kill any RoI potential in LLMs forever. They're playing the wait game instead, observing how the digital sphere reacts to every little increase in price.
If that weren't the case, they'd be pricing at lucrative premiums and even gotten away in short-term considering the increased dependency in the enterprise world. But that'd be like killing for the golden egg too soon and losing all long-term potential.
Once the folks are so addicted to LLMs that even writing a hello world program sounds like a nightmare and coming up with an article draft feels like reinventing Egyptian glyphs, that's when the real pricing hammer will come.
Anthropic and OpenAI won't be around in 1-2 decades if this is their long term plan. People are not going to revert, but go elsewhere. China is proving that it can be done cheaper.
I don't think anyone has a firm grasp on actual inference costs -- including the research and training that has gone into those models. We've got near-frontier capabilities from open source models from China at pennies on the dollar compared to US big tech rollouts. OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
> I don't think anyone has a firm grasp on actual inference costs -- including the research and training that has gone into those models
We know roughly how much these companies spend and what their revenues are. Based on that, they'd have to more than double revenue (without spending more money) just to stay even, and that's not good enough given how deep in the hole they are.
> OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
Both are true. I mean, I'd be willing to spend a bit more than I do now, but not more than double, and neither are most companies. The company I work for is currently investigating how to reduce LLM spend, not looking to spend more.
> OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
Both. They are charging the most they can get away with and that amount is still heavily subsidized by VC capital.
> I don't think anyone has a firm grasp on actual inference costs.
There are huge numbers of users (myself included) that do have an exact idea of what inference costs on open models. Because we can buy tokens from 3rd parties that have no motivation to subsidize our use. That's to say, there's a fair marketplace[1] and we're hanging out there.
If you want to say "I don't think anyone has a firm grasp on actual inference costs on these proprietary/closed models", then I could agree with that.
Just outta curiosity, as I’ve never gotten a spend anywhere near that, what variant were you using? Like max context window and fast mode? Or was it just chugging along non stop for three days?
Fast mode max content window. The task was: replace all 1600+ queries from one database to another and make the whole integration test pass. We did multiple passes, with different concerns when changing from database to another. My OpenCode session right now says $4,365.02.
I haven't gotten close to this either before, but now we wanted to move fast because this branch gets conflicts all the time and we want to get over with the migration asap.
We have a firm grasp on actual inference costs from the various open weights model providers on OpenRouter. They don't have the money to subsidize inference and it's quite a competitive market, so the prices are representative of the costs.
regardless of whether that's true or not, US companies doing hosted inference of the models coming out of China are also significantly cheaper than those from OpenAI or Anthropic
> Nothing is subsidized. Subscriptions are profitable for both Anthropic and OpenAI.
Even if subscriptions are locally profitable (i. e., the cost of the subscription covers the cost of inference), they're still subsidized because they don't cover training and running the company; otherwise, these companies would be profitable.
I can see that being true, and it very likely is true. But isn't infinite VC money and no incentives to optimize operations the reason behind that?
Take a look at China for example - they have no access to NVIDIA, so they're trying to build their own hardware, they have no unlimited funding, so they try to optimize things.
And Anthropic is complete opposite of that - if NVIDIA were to triple their prices tomorrow, Anthropic would still pay them.
In the end, either we all somehow go mad and start paying Anthropic tens of thousands of dollars per month so support this madness, or we will go with whoever isn't lighting cash on fire.
> Take a look at China for example - they have no access to NVIDIA
Not true. Stop following US media spam if needed.
1. Very recently, the US did close a loophole on sanctions that allowed Chinese companies to use NVIDIA hardware outside of China i.e. before that was closed they all had access. The trick was train outside, do adjustments, ship the disks back and use non-NVIDIA in China, but at least the training and endpoints not hosted in China could all use NVIDIA.
2. There's been plenty of reports including fines and bans e.g. to Supermicro on smuggling NVIDIA hardware to China. I doubt it has been stopped. You can't catch everyone.
"Nothing is subsidized" is a wild take. They might be making money on some users, perhaps even most users, but certainly not all. Also, "subsidized" doesn't just mean on compute.
I do, and it's called DeepSeek's pricing table. At the same time, "subscriptions are subsidized" cohort have no data whatsoever, and yet they're in every thread.
Granted, it could still mean that Anthropic just chooses to lose money - but that's Anthropic's choice.
DeepSeek has proven that inference can be much, much cheaper than what Anthropic advertises on their API rates page.
Just a personal anecdote but I have not hit any more thresholds or limits since switching to the MAX plan and so far, it's been worth it. But I do wonder how long even this will last...
I think subscription models are sustainable, but longer term, we should probably expect to see more prompt optimization happening in the providers inference pipeline. For example, unless you explicitly tell the agent or API to use a specific model, fronting the inference layer with a caching prompt classifier to determine which model to use, and automatically select the lowest cost model would probably already save alot of money (IDK if Claude/OpenAI do this on the backend, but several services I have worked on do some things like this to reduce costs of delivery customer facing inference at scale).
> fronting the inference layer with a caching prompt classifier to determine which model to use, and automatically select the lowest cost model would probably already save alot of money
Unfortunately, that doesn't work within a single session. The K-V cache of a model is intertwined with the model's configuration. Switching models invalidates the cache, meaning everything up to the point of the switchover is processed like a new, uncached input token.
Per Anthropic's pricing doc, an Opus 4.8 cache hit costs 50¢/MTok, while Haiku costs $1/MTok for uncached input.
Model selection works best if sessions are short and self-contained, particularly if the first few interactions can reliably classify the model need. That probably covers most 'support chatbot' use-cases, but it doesn't describe the kinds of heavy agentic automation that really chews through token budgets.
> The K-V cache of a model is intertwined with the model's configuration.
I don't think this is true if you simply quantize the model or run it with fewer active experts? The underlying weights would stay the same. You could also play further tricks with skipping some of the model's middle layers outright, which works surprisingly well due to how skip connections are used.
There is a definite financial incentive for people smarter than me to solve the problem, and I don't generally bet against businesses finding ways to reduce costs :)
I have the $100 plan and had almost never run out of credits until I started using the ultracode / workstreams feature w/Opus 4.8..at which point I managed to blow the full 6 hour allocation in like 20 minutes, or so. In fairness, it did some amazing things with the extracted information, but it also strongly suggested that I'd need the $200 subscription *plus* a budget for extra usage.
I have been using both codex and Claude in my day to day, trying to not get to attached to one. I want to be able to work with any provider in case one of them does something bad.
I've found Codex to be the better subscription for OpenClaw, because the limits are indeed very generous. However, I've found more and more that Claude Routines/Scheduled agents can replace all the tasks I use OpenClaw for, so I've been slowly switching over to Claude Code. Aside from OpenClaw, I don't find a lot of value in Codex as a harness on it's own.
I feel like Codex made a big push to run everything on your laptop. With Claude, I get 4 cpu's, a fair amount of ram and 30gb for every one of my dumb ideas for free in the cloud containers. Codex used to be similar, but last time I tried it just kept pushing me to run it locally on my laptop, which I really did not want to do with 20 requests going at once. That's the main advantage for me at the moment.
What runs in cloud containers? The dev servers, builds, etc.? I tried to quickly glance at the Claude website and it doesn't mention cloud containers on their pricing page.
Yes, correct, they both have the same capabilities, however it felt like codex was pushing me harder to use my local desktop in an annoying way, while claude code was happy to spin up a bunch of dev containers for me in the cloud.
I do slightly prefer 5.5 for complex work but Claude quota usage has gotten infinitely better since the dark days a few months back - has gone from being infuriating to something I pretty much don’t have to worry about with it as a daily driver. (In fact, hitting GPT weekly quotas is more annoying now). Understand if people are still scarred by the issues + poor comms around them, though.
The War Department has not existed since the passage of the National Security Act of 1947 and the government department has been known as the Department of Defense under US law since the act was amended in 1949. If you have an issue with it, take it up with Congress.
I've only ever had the $20 month claude plan but last night took the time to setup opencode + openrouter paying for deepseek + glm. Previous experience, while extremely awkward, I'd hit my limit within one or two chat replies and it'd take me like 4 limit cycles to complete my task. Now I'm able to complete an equivalent task entire task for less than $2 in two cycles (ask -> revise).
I'm doing basic web development here utilizing animejs. Nothing too complicated (mostly saving time doing the scaffolding, still write the bulk of animations manually).
Truly believe that American companies are going to get completely curb stomped by China due to greed, ineptitude, and violating the social contract.
Sure, modern American corporations care more about hoarding wealth rather than helping build up US society. Once neoliberalism became the mainstay economic position of the US income inequality has skyrocketed, healthcare costs have increased, childcare is more expensive than university, housing has become both unaffordable + unobtainable. By simply existing costs have increased while life becomes unstable.
Why aren't corporations doing more to help workers with childcare? Why aren't they doing more profit sharing with workers? Why aren't they encouraging unions or sectorial bargaining? Why isn't the government mandating any of this?
Americans very rarely benefit when US corporations do well. That needs to change. No one benefits if Meta continues making billions in profit every quarter while society suffers from isolation, depression, suicide, and scams from their services. Americans don't benefit if health insurance companies are making massive profits while they can't afford deductibles.
Our society has been setup to simply extract wealth in all facets of life. That's a sick society and it needs to change.
I'm not saying China does this better, in fact China has some of the worse worker rights out of all the industrialized countries; but at least American consumers would benefit from cheaper higher quality Chinese goods. The world would likely benefit too if America got off the cold war hype train that did nothing to benefit humanity outside of those making weapon systems.
For me it almost immediately blocked. I had it writing code related to message digests - and it seemed to think it was too gifted for that. Gave the security warning and switched back to 4.8. Whatever... it will probably soon have the API error soon. I have mostly switched to the Codex 200 a month plan. I've found their 5.5 xhigh to be better than Opus 4.8 "ultracode." Also, i have not once seen their servers fail for compute unavailability, unlike Anthropric which happens almost ever hour.
I asked Fable for a complete code review of my programming language code base. Spent like 10 minutes thinking and then switched to 4.8. Burned like 15% of my weekly usage too.
> Fable 5's safety measures flagged this message for cybersecurity or biology topics.
> They may flag safe, normal content as well.
> These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them.
I had a similar experience. I wanted to test it by asking it to summarise a scientific OMICs-related paper. It gave a warning about me potentially developing a bio-weapon or something like that. And switched back to Opus 4.8.
We just blocked it at our org for this reason. They will "retain agent request and output data associated with this model, regardless of you Cursor Privacy Mode setting."
The announcement details it. They're storing 30 days of data on all surfaces, first and third party. They claim it is for security purposes so they can review and check for long term jailbreak and distillation efforts.
They also, FWIW, say that they've instituted new policies on their end such as logging any human access to the stored data and automated deletion after 30 days in "most" cases (with another link to a document detailing that further).
I would not use this if you are on a subscription. In <8min it burned my entire 5hr window (which has just reset it appears, I have over 4 hours till it resets) I hadn't used CC at all today aside from this) and then it used up ~$15 more in usage before I could stop it.
The CLI when you select it says it has 2x the usage as opus. Not sure if that matches what you are seeing.
I do wonder if you switched models mid-session, you would have lost all your cache. Reloading the context into cache can really eat through your usage.
That’s odd, I used it on a pretty complex refactoring task and it worked for 22 mins and used only 15% of my 5-hour limit. I’m on the $200 Max plan though.
Fable seems very good at finding bugs (unsurprising given Mythos lineage), so this seems a pretty smart strategy. Once you see the bugs it finds in your existing Opus code, it's going to be hard to go back, psychologically speaking.
Considering their apparent nerfing of the end user plans in favor of enterprise clients, is Anthropic still the "more ethical AI company" like everybody loves to tell me all the time?
Assuming this isn't just a supply issue on their side, nothing says "ethical AI" like only allowing mega corporations to use it through cost barriers.
You really misunderstand what AI-doom people are worried about if you think this is anywhere near the top (or middle, or bottom) of the list of concerns.
If you can't trust them to act ethically on the small scale, why would you expect that to turn around once it gets to a larger much more important scale?
How many government sanctioned school bombings does it take for them to quit working with said government? For now we know that number is somewhere between infinity and 1.
It literally does not register as "unethical" at any scale to have different products or prices for different customer tiers.
The question of collaboration with USG is a much more complex one, but is not the one raised above.
Edit: I'll also add that I doubt any AI-doom people "trust" Anthropic per se. The entire angle of questioning – again – misunderstands the AI-doom argument. You appear to think that if companies behave unethically, they cannot be trusted and they will not produce good outcomes, inversely: if they behave ethically, they can be trusted, and they will produce good outcomes.
Any competent AI-doomer would argue that ethics or trust are essentially irrelevant.
The entire problem is that people can act totally reasonably, even ethically, and this is not a guarantee of good outcomes. Situations can be created in which completely ethical, reasonable behavior actually produces a bad outcome. You do not need to assume people are bad in order to produce a bad outcome, and inversely you cannot assume that you will get a good outcome from good people.
"Arms races" are one class of situations that often have this characteristic. "Bureaucracy" is another class that we encounter a lot in daily life. There's a lot of them!
Yeah, it's positively precious to think the specific pricing strategy for consumers is the overriding ethical concern with OpenAI, etc. I don't have any particularly strong affinity to any AI company, but comparing pricing to say mass surveillance is ... something else.
Your original comment was about pricing ethics, does Anthropic’s connection to the DoD have anything to do with pricing ethics? They’re in no way coupled, one can be ethical while the other is not.
even for Pentagon thing, Dario said he doesn't object military AI, but said Claude is not ready YET. I speculate he was afraid of reputational damage from cases if Claude would guide missiles on elementary schools.
Setting aside the simple fact that there is no ethical consumption under capitalism, the reality is that regardless of how Anthropic feels, it is becoming clear that many, if not all countries regard AI developments as strategic technologies (and they should).
Anthropic needs to be at least somewhat in the good graces of a capricious administration that is already under pressure from businesses and citizens to regulate AI companies across multiple different domains, whether it's energy consumption, job displacement, military and defense applications, surveillance, etc.
If Anthropic wants to survive, they need to acquire influence with the government that most impacts them as an American company, and a massive exporter of services in the AI space to other countries, otherwise they could get locked down and locked out of the market for national security reasons.
It sucks, but sometimes the survival choice is to make an ethical compromise in hopes that you can still be around to make better decisions later.
> Setting aside the simple fact that there is no ethical consumption under capitalism
This "simple" fact needs quite a bit of additional context and work. Making grandiose ethical claims like this can be countered with other grandiose claims such as the fact that there is no ethical existence under communism or socialism.
It only needs additional context and work if you are unfamiliar with the concepts underlying it. Possibly consider you are out of your depth here, rather than jumping to conclusions.
Sure. Why not, I'm bored today and waiting for some stuff to finish up :D
The fact that there is no ethical consumption under capitalism is not material to whether or not ethical existence is possible under communism or socialism. In order to survive in a capitalist society, one inherently has to make choices that require trade-offs, and those trade-offs are burdened by a history of decisions made not just by the people alive today, but our ancestors as well. Does that mean I walk around chanting "Reparations", "Land-back", or other calls to action? No, but I do acknowledge that there are unresolved issues and as a Canadian, I know we need to do more to resolve treaty issues, and environmental issues, and system discrimination. I also know that Americans need to do better to address systemic discrimination and many, many other issues. It also doesn't mean I want to give back my house, or give away all of my possessions. It just means I try to make good choices and support businesses and people that are open about the trade-offs they make and try to engage as ethically as possible.
Acknowledging those facts doesn't absolve us of responsibility, it's a framework that allows folks concerned about whether or not they are doing the right thing to accept the trade-offs that they choose to make and be responsible and accountable for those choices to themselves or their communities.
We live in a world with scarce resources. It's possible that with a foundational redesign of the global economy, and the requisite authoritarian government that would be required to force such a redesign, we could eliminate food scarcity, solve energy scarcity, and make sure that everyone has a place to live. Those trade-offs are probably not worth the ethical cost in political and physical violence required to accomplish it. We have seen the trade-offs that happen when the powerful are able to exploit communist or socialist governments. We are seeing the "late stage capitalism" impacts of allowing the powerful to exploit capitalism in democratic societies. Acknowledging that the current capitalist system has lead to the greatest prosperity for the upper echelon (financially) of humanity, and a dramatic reduction in global poverty shouldn't obscure the reality that much of that wealth comes from exploitation of people and the environment.
It's a huge problem to unwind, and we can't let the burden of every choice that we make stop us from trying to do better, but we (as in society in general) can't do better if we don't at least acknowledge the compromises we are making along the way, and try to plan to fix it in the future.
Probably a topic better suited to beer and a pub setting than HN though :P
Where is your evidence that this is Anthropic backtracking on its ethical and contractual commitments rather than DOD backtracking on its blatantly illegal coercion (which it's almost certainly going to be successfully sued for)?
As someone that was in Minneapolis during the ICE raids, including one where a US citizen at a nearby restaurant was thrown in prison for 3 days despite having his passport on hand because he looked asian, it's hard for me to not equivocate the ethics of AI companies actively collaborating with the Trump administration as different flavors of ice cream.
I don't think offering a product under a certain set of terms obligates a company to maintain that offering forever. The bait and switch is certainly annoying but seeing as they're very upfront about it you can't say you weren't warned. Don't like it? Don't use it.
Yup - who cares about x-risk or red lines for domestic mass surveillance anyways? I draw my red lines at prioritizing profitable customers when heavily resource constrained. That's the true definition of evilness!
More of a free trial to those authenticated and qualified with existing payment. Subscription billing is going away for sure though eventually based on the economics. Token “all you can eat” is a capital furnace otherwise.
(I’m highly confident open models will eventually achieve a similar performance benchmark with distillation over time)
Subs lose money on individuals to get those individuals to force their companies to pay for the corporate plan. The economics are bad, but so are the economics of grocery stores selling Milk and Bananas at a loss to drive traffic, which they basically ALL do.
I pay a lot but barely use it except for some intense days, where the lower plans would have throttled me in like 30 minutes. API billing is still more expensive. If you want to not pay much, go to openrouter and use chinese models. They are cost efficient.
Retain and hire the engineers who don’t require heavy use of AI to deliver value? The current SWE job market speaks for itself. Where will you go where they will let you burn up tokens in a high cost of capital macro?
ZIRP (zero interest rate policy) is over, software engineers no longer call the shots now that there isn’t vast amounts of capital chasing yield, and that capital bidding up salaries and keeping the labor market for engineers tight.
If you are x more productive with generative AI, very shortly you are going to have to prove it with a token budget (or, if you’re lucky, an org willing to spend for on prem hardware for capped token cost, fixed capex vs uncapped opex).
The comparison is not SWE vs SWE with AI. It is SWE vs SWE with AI with a constrained token budget ($x/month) delivering the same value at the same or lower cost. If you cannot prove that you are wildly (vs marginally) more productive with the AI, why would they pay for it? Prove it.
That's a big problem for all of the AI companies. Most people don't find the technology compelling, accurate, or ethical enough to pay for a subscription.
Why wouldn't Anthropic just wait until people start subscribing, do some kind of marketing push, or obtain some kind of other sustainable revenue stream, before they go IPO? I wonder if they see the writing on the wall with all of this and want to cash out as quickly as possible?
That's not how it works. They don't need revenue, they need addicts.
Specifically they need businesses that fired people and adapted their business to the products, so when the unsubsidized costs hit the businesses are forced to eat the true costs.
Yes they can't afford to give the products for free, but what is essentially happening with AI services is economic dumping, keep costs artificially low to get people to fire everybody, and then Jack the rates once they have Monopoly control
But the only companies firing people (and certainly not everybody) are either the companies with an AI or the investment and finance firms that stand to profit from AI. I smell hype. And no company is firing everybody because of A.I.
I agree. They need addicts, but they are high on their own supply and everyone else can see the danger in getting hooked.
I just assume Opus is constantly nerfed based on capacity. I was exclusively Claude for a long time, but the inconsistency in quality, constant outages, and slow downs were too hard to work with.
I just use dumb and fast models now. I'm more engaged. I think that the higher the quality of the model, the more you tend to vibe with it, and then the more hallucinations you then miss. I'm not sure which is more productive, but I definitely burn out faster the more I vibe. At some point you're spending your time on forums, discord, or youtube instead of engaged with what you're building. Or you yak shave about your tooling and end up creating the 600th multi-agent gastown harness and blowing thousands of dollars on tokens to create it only to discover it's too expense to actually use.
I agree with you. The more I vibe code, the less interested I feel in what I'm building. Working with models that force me to think, especially with personal projects, helps me stay engaged and enjoy what I am doing more.
It's possible that they will transition to usage credits but why not take them at their word? To date they have continued to offer better and better models to their subscription plans.
Upd: I meant big picture, not with respect to this model release. Where do subscriptions figure into their strategic vision. Will consumers end up paying enterprise prices in the future?
In the blog post they say when sufficient capacity allows them to do so they aim to restore Fable 5 as a standart part of subscription plans and intend to do so as quickly as they can.
Even Opus 4.7 felt like a regression from 4.6, consumed a lot more tokens while I didn't experience any substantial improvements. The company I work at simply rolled back to 4.6 on everyone's configurations, disabling the toggle for 4.7.
HN needs to take a chill pill. Could it be that Mythos is expensive and they just want to give people a taste of it? I mean the alternative is not offering it at all?
This is just the sales team doing their thing, applying the Law of Scarcity to drive demand.
It's the same exact speed as opus >=4.5, sonnet 4.5, and twice the speed of opus <=4.1
It must have about the same active parameters, or else its a larger model running in turbo mode (smaller batches) and being heavily subsidized for some reason. But given most of the benchmarks are within 5% I doubt it is a much larger model. Most perplexing.
I don't think they'll phase out subscriptions ever, their whole play has been to drive demand from the bottom up. Get engineers hooked on building with claude at home, then get them to demand the ability to use it at work, and bend over their employer with no lube.
They'll probably tighten the quotas to reign in whales though.
They almost certainly already make a fuckload more money off API pricing than they do subscriptions, even if there might be more total subscription users. So offering subscriptions even at some loss is probably going to continue. Honestly, I'd be surprised if they even lost money on most subs; there are definitely Token Whales out there who mess up all the accounting up, though.
Realistically I think Anthropic just has insane demand but finite capacity to run models, and Fable will just make them more money if they dedicate it to API pricing. I suspect the goal here is something like: get individual engineers/PMs on their personal plans to taste Fable and then go to their meetings and say "Yes doubling the price of every single input/output token is a good idea, boss".
I certainly hope not. PAYG is not predictable enough for smaller companies or individuals. Where I work (non-tech company), PAYG would never fly. We aren't big enough for that. Of course, you can set usage budgets, but there's a pretty big difference between $200/user/month vs. the equivalent PAYG usage being closer to $1,000/user/month, if you currently use the subscription plan to its limits each week.
Going PAYG only will effectively take these tools away from a huge amount of people and accelerate the push for local LLMs.
OTOH, accelerating the push for local LLMs would also be fine with me.
I doubt it, given the importance of those subscriptions for building and maintaining market awareness.
The AI landscape is changing rapidly, and with Apple announcing the option to change the AI backend, and potential requirements enable AI choices as well, similar to EU browser choice requirements (this is more reading tea leaves than any actual requirements I am aware of). The new OS changes coming to support Googlebook, and deep Copilot/AI integration into Windows will make maintaining user facing subscriptions essential for independent model developers like OpenAI, Anthropic, and Mistal to remain relevant longer term.
If the don't maintain that relevance there is increasing likelihood that they will get consumed by other companies whether it's Apple, Microsoft or Google to form a foundation for their OS, or other cloud providers.
That make sense, but what about the specific bifurcation we're seeing here of super primo models versus still good models being available to subscriptions?
It's kind of annoying not getting access to the primo model and paying 200 bucks a month. I understand 200 bucks a month is basically nothing though.
Like I don't totally understand why they'd let me have it for a couple weeks and then take it away and say I can have it but I have to pay retail and retail is like $1,000 a day.
It's better to have loved and lost than to have never loved at all??
It's a trade-off. Every hyperscaler is buying and building compute capacity as fast as they can dodge red tape. There is limited compute capacity, and scarcity is a real thing.
As a consumer I can choose to buy subscriptions to a range of things, including $5 droplets or VMs on a broad range of cloud hosting providers. I can even buy cheap bare metal at a bunch of providers at an affordable retail rate.
I can also buy "unlimited" AI packages that will be optimized to fit the cost model from a variety of services, with different impacts, such as rolling outages when I consume a daily or hourly allotment.
Right now VC and the investor class are subsidizing the rapid evolution of the services and availability, but that VC is running out. In more traditional economies, AI would have developed and rolled out more slowly, and through metered subscriptions, with the eventual rolling out of "unlimited" packages like telephone, internet, or cell services once the market became commoditized.
We have seen a big inversion of that with the race to "win" AI marketshare. Now the true cost is being exposed, and the most competitive and capable models are hideously expensive to operate, so it makes sense that we are moving to metered billing for a utility service. If you want gas, you can buy regular or premium. If you have a premium car you definitely want the premium, but for most people regular is good.
Give it a couple of years, and the survivors will settle around fairly industry standard models of consumer grade services, pro-sumer accounts, and business/enterprise models.
Things are still shaking out, but I get the sadness. Luckily I work at a big tech company who is banging the drum on doing experimentation so I use my prosumer claude pro and other accounts at home for hobby stuff, and save my heavy lifting and potentially experimentation for work :P
"we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design).
...
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user."
i doubt that's the goal for them. i bet they just really don't have capacity for people using it a ton, yet they wanted people to be able to try it out while it's new. so they compromised and made it temporarily available. and then hope they can get costs down or capacity up so they can make it more available again
I think the goal is "private citizens: subscriptions; corporations: per-token billing." It's getting people addicted to LLMs on cheap subscriptions so that they can then force companies to pay for expensive inference.
I’m just about ready to cancel my small business 5 user plan with max licenses, because although cowork is really great. I just find OpenAI/Codex to be a lot better most of the time.
> The "offer, then remove" aspect is a bit eyebrow-raising -- it feels like they are trying to get subscribers to switch to usage-based billing, which makes me wonder if we'll ever get it after that June 22nd window.
My guess is that it is a massive model similar to GPT 4.5 and $10/$50 pricing is for its output will discourage people from using it. I also read safety = nerfed.
This serves as a good reminder that relying on AI models is borrowing your tech from someone else. They can take it away or raise the prices arbitrarily.
If you rely on this as a core part of your business/profession, you will be at their mercy and subject to whatever whims or challenges they have.
> Pricing for both models is $10 per million input tokens and $50 per million output tokens.
The step-up in intelligence looks massive (we'll see in practice), but the price is getting to a point where it's making me question if it's even worth giving it a try.
Good competitors will probably be out soon, which should level the playing field. I am more excited about that, just the fact that they showed that such an improvement is possible. I'm okay waiting a bit longer for this to become attainable for plebs like me.
Sadly this does not seem to be the case here: if you read the announcement entirely, they include a "cost per task" metric which basically continues the trend of their previous models. So yes, tasks will cost you more, but results will be better - allegedly.
I don't see how it won't be. They lose insane amounts of money on subscription plans. I'm sure they still lose money on usage-based billing, but probably not as much.
Most gyms sell more subscriptions than they can fit under their roof at one time. If a gym only sells to heavy users, it will either be constantly turning members away or have to buy more equipment. Its equipment will wear off faster. Depending on amenities, it will go through towels, soap, water, et cetera faster, too.
It depends on the gym and their business model! A super-budget gym like Planet Fitness that charges $15/month is going to lose money on heavy users, but they count on most of their members being infrequent gym-goers. A luxury gym like Equinox that charges $300/month can target heavy users without any issues, and they'd actually rather members go more so they stay and spend money on expensive salads and smoothies.
Right now all these AI subscriptions are priced like Planet Fitness, but they're used like Equinox. They're hoping that the new a la carte offerings will move their pricing more in that direction as well.
That doesn't mean the company is losing money in aggregate on these subscriptions. Buffets are still in business even though some people gorge themselves silly at them. The incremental cost may exceed the incremental for a particular person or minority group, but that's not how these businesses measure profitability.
I assume consumers aren’t a big note in their bottom line. I’m not actually very sure about that, just an assumption.
What I wonder however is if these tools will become something I use at work only. $100/month is already a massive stretch budget wise. If these models keep devouring tokens there’s no way I’d get the same usage time out of them for $100 in usage credits.
I just don’t think I’d use them much at all at home.
If you think about it, the more people pay for these new and more resource hungry models, the longer it takes for them to become no extra cost and the longer it takes the more people are tempted to pay extra.
But they're behind by quite a bit now. CFO (of OAI) Sarah Friar said the next training run will be in the fall on Vera Rubin, I think that means I'll have to wait > 6 months?!
> * On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.
Of course, they are a casino as well giving you free spins at the wheel with their new Fable machine, and it is done on purpose.
Once there freebies have expired, many of its users will begin to gamble more on the new casino machine and will realize that it is expensive.
It’s an interesting thing to bring up because it’s this classic thing we’ve seen for decades now.
The ramifications go beyond the individual which is why I assume they mentioned it. They don’t need to use it/not use it for it to have interesting implications.
The pelican has looked very same-y across all frontier models, same color bike, same camera angle, etc. I suspect this challenge is already too embedded in the training data to be a good signal when it succeeds, and maybe even when it fails in pathological ways mirroring existing AI pelicans on the internet.
SVG generation is a good test because it's extremely easy to subjectively assess with visual reasoning where humans are strong. However, pelican on a bike specifically may be overused at this point.
Variations of this comment have been posted for over a year. The pelican has now morphed into part of HN culture rather than a legitimate benchmark, but it's still valuable as a meme.
I'm beginning to wonder how much of a useful metric the pelican is because surely the frontier labs must be training their models on pelican-artistry because of how well known your test is now?
Simon has addressed this on virtually every new model release. He also has unpublished alternate prompts. But the larger point is: this is a fun experiment, not a serious and objective benchmark.
It's silly and a joke and a surprisingly good benchmark and don't take it seriously but don't take not taking it seriously seriously and if it's too good we use another prompt but don't actually because then it's not the pelican post and there's obvious ways to better it and it's not worth doing because it's not serious.
Only coherent move at this point: hit the minus button immediately. There's never anything about the model in the thread other than simon's post.
But what if they are better at flamingos? Are they optimized for pelicans? How about “draw me a four headed owl”? The meme, I get it, but I’d settle for a working bash script, tbh.
I just run my own benchmark for "draw an SVG with $animal driving $vehicle". I won't post my choice of animal and mode of transport, but there are plenty of uncommon combinations to choose from. So far it's a fun and visually intuitive benchmark that does seem to correlate with model capabilities
I don't know. Just looking at the bike frames (specifically the fact that the AI generated bikes have rather unsteerable front forks), it's clear to me that frontier labs aren't spending much time tuning models to make bikes look coherent, which I assume is an easier task than making a pelican riding a bike look coherent.
I've seen this reply to Simon's benchmark for 2 years running now, and yet you still see improvements and objectively-bad results over time from new releases, even when I'm sure every frontier AI team has/had a person at least partially dedicated to better bicycle-pelican SVG outputs. Alas.
I've been enjoying seeing how the quality of individual models differ based on the amount of reasoning effort you give them. If they were baking an a good pelican you wouldn't expect them to differ so much.
I honestly assumed their comment was tongue in cheek humour, because positively no one actually cares how these models generate an SVG pelican riding a bicycle. It's some meme thing that this stuff always appears here.
I also look for this reply because i like seeing the follow-up reply saying that this is not a benchmark anymore because labs have gotten it in their training data.
that reply never failed to come it's basically a meme at this point
I don't really consider that a great benchmark anyway and we really need better ones that are objective instead of these mostly performative and cheatable and also available in the training set.
I think it's a clever thing he did to basically guarantee he continues to get major traffic to his blog here every time a model is released, especially since he's taking sponsorships with a static banner at the top of every page now. I think he's trying to go the Daring Fireball route.
> Pelican for Fable 5 on default settings is a clear improvement on Opus 4.8
And doesn't contain any actual criticism within the comment (your blog post might, but just referring to what was posted on HN, which is a bit booster-y on its own).
The entire pelican benchmark is a joke. The joke is that, for all of the billions of dollars poured into these things and the claims of PhD level intelligence, they still draw pelicans not-much-better than a five year-old would.
I don't spell that joke out in every comment I post here because that wouldn't be very funny.
It seems like Fable will refuse to do any work when it comes to developing LLMs or even asking questions about topics related to LLM. Simple things like asking to explain a paper fails!
From the model card:
In light of the ability of recent models to accelerate their own development, we've implemented new interventions that limit Claude's effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design. Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms. Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user.
• My most noticeable immediate jump was in how its frontend design was much more intentionally crafted, and delightful without feeling like 'AI vibe coded'; with better end-user usability too.
• In some internal agentic harnesses, it achieved better results with about half the tokens, making it cost the ~same as Opus 4.8 price-wise! The real price increase is less than 2x; with biggest differences in harder problems where Opus 4.8 struggles (or needs many turns).
• Part of the token efficiency improvements come from Fable doing more targeted and surgical diffs, with less non-necessary changes. This is great, because PRs often have less LoC changes for review. It writes more maintainable code without explicit human steering.
• For general conversation and assistant style use cases, didn’t really notice a difference vs 4.8.
• 1M context window, without increased pricing for long context is AWESOME. This is a massive win.
• The classifiers are super aggressive and sensitive and this does happen for very benign, non-security coding tasks. Fallbacks to 4.8 worked like a charm; but the filters are definitely super sensitive.
Overall, I would describe this as a step change and worthy of the "Claude 5" model name. It did take some time to understand the intelligence ceiling of this model; and even with an extended testing window I'm still discovering new things and often surprised (in a good way) by the model.
I just ran it on a tough reverse engineering problem I'm having that neither Claude Code 4.8 or ChatGPT Codex 5.5 could figure out. 30 minutes later Fable has it all figured out perfectly.
"incredibly" is doing a ton of work here. I do not think its doing even moderate work on visual design, but it can spew out a lot of ui that looks arranged ... ok.
This is still not in the range of shippable UI for top end companies. Maybe for internal tools and enterprise.
At our comapny we limit to protoypes at most and even find it limited there.
Yep. We have some interesting problems, like getting LLMs to create/edit Canva designs in our own proprietary format, which isn’t published or documented on the web. So the model has to work with it, purely from a very detailed system prompt spec / in-context learning.
I assume it might be a good barometer for generalised intelligence; esp in the visual space.
> In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations
It's bad that Anthropic can determine what this means. If you're building a modern app you're likely training your own embedding models and now anthropic can just silently sabotage your training pipelines?
>We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations
At the scale of API requests that Anthropic sees, I think the affected organization count might be substantial, and they might not be getting the full model capability that they're paying top $$$ for.
AI vendor’s idea of safety has always been safety for the interests of the AI vendor in question. This is not a new development, though this may help more people realize it.
How do they detect whether an experiment being done on a smaller model is used to improve a competing frontier model, or just an innocuous hobbyist LLM experiment?
infering the surroundings, like everything else. they will probably look at which company is your email, and if you wrote "better than claude" on the readme.md
this is LLM, it's not like a science or something.
3 months before asking for what to eat before a linear algebra exam trips the machine learning topic ban is my guess. I got flagged immediately asking why my JEPA thing breaks weird.
Not missing the forest for the trees, this effectively means in 3-5 months China will drop open source models that are every bit as capable and dangerous as current day Mythos except with no safeguards.
And the only companies safe from this are the large corporations that shook hands with Anthropic? Because Fable doesn't seem to have actual safeguards, more like 'if you talk about this you will be talking to Opus.' It doesn't guard against offensive use, it prevents all use (offensive AND defensive).
Rationalists are inventing oligopolies from first principles, absolutely incredible things happening in SF
My bet is that Mythos is still over-hyped and the cybersecurity fear and guardrails are mostly marketing to force company partnerships through Glasswing and get public attention.
People quote the "GPT-2 is too dangerous to release" thing as if it were wrong, but given all the slop all over social media and how it's used to create division and attack social cohesion, he was clearly right.
I still remember it. "Open"AI going API-only because GPT-3 is really really dangerous, so forget the Open in our name and all of that, you can't download our models anymore and must request access to them because they pose a THREAT.
Fast forward to today and GPT-3 has laughable performance.
Even back then there were plenty of people who got fooled by AI generated articles. It's easier to spot AI writing now because we are so used to it. They were right to be concerned; not that it achieved much since oss models run laps around gpt-3 now.
But it seems like that was not genuine concern, but instead a tactic to pivot to closed models and an API service with an excuse to do so, breaking the public's expectation that they would be a non-profit making open models, like their name implies.
"We had to do extra work to make this safe because it's so advanced and dangerous..." how many times can they trot out that line before it loses its effect entirely?
I mean, they do actually describe what that extra work was, and people elsewhere in this thread are complaining about the effects of those safeguards. So it's not like this is purely empty rhetoric.
people are not questioning whether they did the work, they are questioning whether the work was really necessary (i.e. if mythos is really so good that it needs safeguards to prevent malicious actors from using it)
Yes, and "in collaboration with the U.S. Government" feels like a very gross ploy at appeal to authority. You don't need Mythos or really any SotA frontier model to make malware or do extensive penetration testing/reconnaissance already. Sure, Mythos might be faster/more efficient, but the cat has been out of the bag for awhile. Even the terminology "infrastructure providers" practically screams "Enterprise leads".
I know a security researcher at Google with access to Mythos. He says it's the "real deal" and that "there are career plans I had that are no longer viable".
It's not even very usable... I tried 2 different chats and both eventually got stopped due to the safeguards
One was a piece of code I gave it to improve, it did so and then started writing tests, some of which tested security so the safeguards triggered
Another was one of the cryptography puzzles I use as new model tests, which are hard to oneshot and there's no public solution anywhere, it completely refused to even try to solve it
They're trained in a model class likely in 2t to 3t range. It's very unlikely that chinese labs have access to gpu systems capable of training models like that, let alone serving them. This requires proprietary room-scale systems which fetch a huge premium over typical 10 slot systems.
I am sure that they can develop their own equivlient version of such clusters in around 1 year though. Distilling fabel 5 will also go a long way.
MoE experts were likely trained independently / in a sparse format. Training anything beyond 2t on typical systems would be infuriantingly slow, you could do 4t on nvidias room-scale solution, but for a reasonable training speed / batch size it caps around 3t.
concept is similar to how it works in inference, instead of performing regressive writes to the entire model you run the whole model, but part of the model can live in system memory and get swapped in/out on demand. So only XB parameters are active in training.
edit: I am not really sure if it works like that. I haven't looked too deep into deepseek v4 pro specifically.
I wonder if model distillation will continue to work as well as it has. Given hidden reasoning, the ever expanding number of expected capabilities, a serious compute shortage, the looming possibility of model collapse, and dramatically higher API costs I would guess that it's getting much harder to do.
Asian labs generated synthetic datasets from UBS labs but also innovated with technology. Now it is harder to get the thinking traces AND Anthropic is recorded to poison it as well.
Thus Asian labs will have to generate their own data sets, which with the huuuuge usage boom from deepseek, mimo, kimi, etc, they will be able to.
Isn't that a good thing in a way? If everyone has the weapon and defense at the same time, we will fix security holes and live safer lifes instead of having some three letter agencies and military backdoors in everything.
Pandora box is open anyway. It's better now for everyone to have the same power rather than a few national states.
Not sure this holds, sadly. I spent a few months reporting serious security bugs as model capabilities took off earlier this year, and only ~half were fixed. The unfixed bugs were just as critical as the fixed ones; sometimes they were even two similarly critical bugs at the same company, and only one would be fixed!
On your other point, the government still has systemic leverage and can compel access, so this doesn't remove that risk.
That doesn't mean this is the end of the world, and some balance of power is usually good. But I do think it will still increase the capabilties of rogue actors and their net harm.
I think we're about to see a big relative drop-off of open models vs closed. I don't think there'll be an open model that competes with Mythos for ~2 years.
Even OpenAI and Google are struggling to get this kind of performance. If the distillation defenses are any good + chip controls prevent China from training massive models, it's over.
3-5 months is a long time and they are pretty useless on arrival because the frontier models are so good, that it's hard to go back even if it's way cheaper. Your work flow is adapted to that level of intelligence for months.
I don't think China has any incentive to arm the rest of the world with highly capable models that can be used against them. Undoubtedly they will continue with the arms race, but they will preserve the best stuff for their own use.
I think the stronger incentive is undermining/undercutting the Western AI companies. Given what we have seen, any model can be used/convinced to do harm so that is just part of the game
I agree, depending on how much of this is marketing and how much is actual capability. It's one thing to undercut models that finish writing assignments for lazy students. If this actually identifies vulns and writes exploits, or if it designs bioweapons, those are pretty different. Those are actual weapons, and I don't think they're going to arm the adversary.
Are the sibling comments astroturfed? This seems like such a bizarre thing to be talking about in relation to an Anthropic model release. As someone from the UK, I don't feel like I'm living in an authoritarian country. And yet most of the sibling comments are insinuating that I am. Weird.
HN is extremely pro free speech and the UK has recently decided to engage in censorship. Part of the issue users here reckon with is the recency. Unlike many authoritarian countries that seem hopeless with regards to free speech the UKs censorship is a recent development that many think can still be undone through political action. Similar to takes on why Israel is being protested when places like sudan arent.
The UK has a censorship bureau, ofcom. The example that comes up most here is 4chan, which the UK is currently trying to ban because they refuse to do age verification. If you read the threads here you will see other stories. One that sticks out to me is someone who was talking about their struggles running a forum about depression. They live in canada and were contacted by ofcom demanding the forum add age verification, cant totally remember the reason but it was something about kids being able to access talk about depression. Ofcom said that if he doesnt add age verification to his forum he will be arrested if he ever enters the UK. He even blocked uk IPs but they said that wasnt enough. We can quibble about whether age verification is a form of censorship, I think it clearly is, if only because it is a large regulatory hurdle that stops people from hosting forums because its too much regulatory work.
They’re talking about British hate speech laws. They think other countries have universal free speech and they absolutely do not, but for some reason they think Britain goes too far. Although “think” is probably too generous - they’re parroting talking points.
> In the UK you get thrown in prison for making a slightly unfriendly tweet.
Do you? The closest thing I can think about is how someone was jailed for encouraging arson attacks on asylum hotels. I'd be extremely surprised if the US had zero cases of somebody receiving a police visit after threatening to kill the President or bomb a school or something...
(FWIW I do think the UK needs stronger free speech protections, but saying that you'll be immediately jailed for writing unfriendly tweets is a huge stretch)
>the quality of discussion on HN has gone to shit, i miss when model released used to have actual informed takes from people that used them or substantive discussion about the system card
[Mythos 5] does sometimes still engage in reckless
or destructive actions in service of a user’s goals,
and our interpretability analyses indicate that it
is aware that these actions are transgressive while
it engages in them. As with Opus 4.8, rates of
evaluation awareness and reasoning about being graded
are significant, and not always verbalized; we
introduce new and more detailed measurements of the
nature of this awareness. The reasoning text from
Mythos 5 is somewhat denser and more difficult to
interpret than that of prior models, containing
more jargon and difficult language.
So, it (often) knows when it's being tested while hiding that fact, is willing to break rules, is great at hacking, and it's getting harder to understand what it's thinking.
Humanity has plenty of catastrophic risks to deal with already, I wish my field was not working hard to add a new one.
Not the direct person you asked, but my answer would be alignment, interpretability, and policymaking. Perhaps improving existing usage? Helping grandma create reminders doesn't require advancing the AI state-of-the-art.
They are state of the art at all 3! As are other labs. Of all the labs they seem to take alignment and interpretability the most seriously to the point where they are hampering their own revenue in service of trying to not cause problems while also being in an incredibly competitive space.
All AI companies are trying to do all of what you’re saying. The issue is you can’t do that for long without a frontier system. Or you become a completely different, far less profitable company.
Implied in my answer was "and not creating ever stronger AIs", which unfortunately the big 3 labs are failing at. And they might be hampering their own revenue by doing the rest, but they also know that rocking the boat too hard is even more dangerous for their revenue. I wouldn't call it selfless.
No it’s not selfless, but I can’t imagine a more shareholder minded CEO would not have done a slow rollout of mythos. The point is: creating ever stronger AI systems is what these companies do, it is integral to what they even are. If you think that’s bad, even if all frontier labs agreed with you, you’re in a horrible game theoretic position. Any player can gain an enormous advantage by breaking the agreement. Not to mention Xi would be absolutely thrilled; now China can take over the AI race, become the load bearing infrastructure of humanity. We live in a complex world where simple childlike ideas like “well why don’t we just stop developing AI” actually are more damaging than keeping things going.
“Alignment” as a goal always ignores the “with what set of interests”, because there is an attempt to maintain ambiguity for different audiences (particularly, users, and non-users who seem themselves as the arbiter of broad social norms) to read in their own interests, when the actual answer is always the interests of the actor pursuing “alignment”.
Which value system to align to is absolutely the right question both rhetorically and otherwise. These models have a fairly western bias due to the domain of the training data.
But also, these models are capable of adjusting their value system depending on the user. Not saying that’s what’s being done but at a technical level that’s fairly straightforward, though not obviously better or with less problems.
Probably MistralAI or any of the Chinese companies that aren't throwing billions down the drain while American society lacks healthcare, childcare, and good wages.
American society has higher wages than almost any other developed nation [1], so it's objectively incorrect to say the US doesn't have good wages. It chooses to make you pay for private childcare and healthcare, both of which are high-quality but stupid expensive. It's a tradeoff like anything else a nation/society creates and prioritizes.
No idea how that connects to the idea that Mistral or DeepSeek are somehow the "good guys" though?
You want Anthropic to fund your healthcare or something? Also, have you seen the impact of these models on healthcare? Also most of our GDP growth this year is from AI buildouts, would you rather that be negative?
And not even considering: Chinese AI companies are the good guys???
It's a five horse race between Alphabet, Meta, xAI, OpenAI, and Anthropic.
Alphabet dropped "don't be evil"; Meta's CEO called their own users "dumb fucks" for trusting him and also clearly thinks "super-intelligence" is just a buzzword given how he tries to sell it; xAI's model called itself "Mecha Hitler"; and OpenAI's CEO was temporarily fired by the board for a lack of candor.
It's very easy to be "the good guys" with this competition.
It's the "If we don't, someone else will" effect. So long as there are competitive markets and competition between nation-states, a single player cannot unilaterally defect from the race, no matter how dangerous it is. Half the comments on HN lately are "wtf Claude is so dumb compared to Codex; I'm switching"-- nobody can slow down while those exist.
We, globally, can stop it. It has worked (so far) for nuclear disarmament, and could work for training large models. I know that policing the usage of computer clusters is not a popular opinion in technical forums, but something has to be done.
Specially when talking about potential superintelligences. And if people think that's impossible, remember that current models would have been considered science fiction just a few years ago.
I don't buy the superintelligence package, but I think uncritical LLM adoption poses plenty of threats to things I care about, in a mundane human-scale way.
Anyhow, I think you're (absolutely! ugh) right about the politics and I try to make the same point to people: whether you love or hate LLMs, accepting the "inevitabilism" framing is just ceding control of the Overton window. For better or worse, technology adoption can be and has been slowed by politics. We don't have nuclear plants everywhere. We don't have Project Orion starships colonizing Mars. We still have very strong social stigmas against genetic selection for human embryos, etc. This all can change in a heartbeat, and I'm not sure that policing the hardware rather than holding specific humans accountable for bad LLM outcomes is productive, but fundamentally: yes, we can stop it.
It's the same deal as Quantum Computers breaking crypto. Maybe there's an 80% chance of it never happening, but when you multiply that remaining 20% by the potential impact...
It hasn't worked for nuclear disarmament. We live in a world where many countries have nuclear arsenals. "But it hasn't killed us yet!" Yeah sure, it's only been less than a century since they were invented. Who knows when nuclear war will come?
True, but look at nuclear tests. There used to be around 50 tests every year, for decades. Now the only nuclear tests in the last 27 years were the six done by North Korea[1]. And there's still only nine countries with any nuclear weapons, and none in the past twenty years[2].
That's a bit better than just "it hasn't killed us yet". I think it shows we can at least stop the further development of this kind of technology.
Nuclear tests are extremely easy to detect worldwide, and enrichment activity is a major industrial process that is also fairly easy to track given the specialized equipment needed.
AI development doesn’t have any of these characteristics. It would be almost impossible to easily distinguish a datacenter that is working on AI development and a datacenter mining cryptocurrency.
It would not be nearly as easy to stop AI development as it is to stop nuclear arms development.
To the extent nuclear arms control works, I think it's only because nuclear weapons are so hard to build-- uranium enrichment is hugely expensive and complicated, and plutonium weapons need actual reactors.
If it was possible for ordinary companies to build nuclear weapons, and also release open-source ones that anyone could use to compete with the paid ones, I suspect we'd all have been dead a long time ago, arms control treaties or no.
Even the (SOTA LLM) open source models are trained with huge clusters. Datacenters are also hugely expensive and complicated.
Or you can take one step back and look at chip allocation. As far as I know there are only three companies on the planet that can make the chips that go in those clusters. One (ASML), if you look back the supply chain to the Extreme Ultraviolet Lithography Systems.
If politicians decided that no more large language models should be trained, it sounds like we could do it.
I think the standard answer is "yes, the consequence of noncompliance is bombing the datacenters, but it wouldn't happen because China also understands why we shouldn't build it".
In 2023 there was an open letter titled "Pause Giant AI Experiments", signed by almost all the big names on the West. I'd say the public opinion only got worse since then.
The system card is 319 pages, at what point do we call it a "book" instead of a "card"?
There's a quote from a METR report on page 52:
>We ran [Mythos 5] on 38 of our hardest software tasks, including tasks centered around R&D. [Mythos5] generally outperformed an early checkpoint of Claude Mythos Preview in these, including by succeeding on some tasks that had not been solved by any public model we have previously evaluated. However, we still observed the model occasionally failing to correctly interpret nuanced instructions in difficult tasks... Based on the available evidence, we believe [Mythos 5] is likely unable to fully and reliably automate R&D for frontier projects spanning multiple weeks. We believe that a better, more confident assessment would require more time, evaluations, and information from the model developer.
Depends whether "unable to fully automate" means "needs occasional human checkpoints" or "slowly stops caring about your actual goal." Pretty different.
That blog post really makes it look like it's graded from an LLM's estimation of an OSS maintainer's review. I see three issues:
1. That estimate could easily be wrong.
2. That estimate is, of course, usable in RL training. This isn't an inherently bad thing, and this is more or less what has improved coding models so much lately. But it does mean that other companies could and surely will do this sort of training, and Anthropic probably did too.
3. OSS maintainers are far from perfect, and there's an unfortunate uncanny valley-like effect in which a coding model can produce code that is just convincing enough to pass review even though it's actually totally wrong. I don't know whether this is a specific issue here.
Given it was made by cognition (team behind devin flop) who now just got to wait out until claude and gpt5 basically do all of the work for them - not very. When you read about it, the framework is highly subjective. Which very quickly becomes a problem because its based on heuristics that probably change a bunch with a better code model.
i worked on one of the benchmarks typically found in new model releases
this benchmark looks very good from the methodology. a cog researcher checking the data themselves is very high signal (not scaleable so don't take the benchmark as gospel, but directionally good)
It's a relatively new benchmark but from what I can tell it has serious cred behind it. I assume it will be picked up as part of the standard suite of CS-related benchmarks soon enough.
Cognition did well in documenting their approach [1].
TL;DR - they worked with OSS project maintainers to build tasks. They score models based on whether a PR is mergeable. All tasks are graded by a human researcher. SoTA models have hill-climbing to do which raises the bar and inspires confidence. I'd say it's legit.
People game benchmarks for fake internet points to get their favorite web framework to the top of the list. I'm pretty sure they will do it for billions of dollars.
I'm not familiar with model pricing trends, did they clearly state how the new pricing compares? (Note that I'm actually asking a question, and am not arguing)
There a few benchmarks out there where all existing models have abysmal scores. So it's not actually a problem if Antrophic's older models are bad, especially if the jump to the newest model is huge, and the competition is also way below it.
Huh? It's a benchmark by Cognition which (1) is building their own models and (2) offers all providers and thus has an incentive to avoid hyping up any one too much.
From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.
After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.
This seems like the pharmaceutical method of get them hooked on the drug with free samples, then once they can't live without it, raise the price. I'm not sure I want to start using Claude Fable on a max plan if it's just going to go away on June 23rd.
But maybe the more charitable reading is that they didn't have to offer this model at all on those plans and they are giving the standard free trial.
It's interesting that we're seeing these gains when it seems Mythos/Fable is "just" a scaled up version of their existing architecture[0].
When GPT 4.5 launched, the gains compared to the model size didn't seem that great, leading some to believe that the only progress we'd see would come from RL.
This model certainly has quite a "substantial amount of post-training and fine-tuning", but it's also based on a new pretrain[1][3], which given the cost, indicate that it is in fact quite a bit larger than Opus 4.X.
[0] One of the early testers mentioned: "As far as I can tell from talking to people internally at Anthropic, there's nothing special about architecturally"[2]
[3] There were rumors going around when Mythos was first announced that it was the first 10T parameter model, but I can't find a verifiable source for that number.
> A new data retention policy
Finally, we’re making a change to the way we handle business customer data for Fable 5, Mythos 5, and future models with similar or higher capability levels. We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases ...
Very interesting. I am not sure this will comply with organizational policies and standards protocols (HIPPA etc.,)
This makes it an instant non-starter for probably 95% of organizations. A lot of people are about to get in trouble for using it before realizing this.
At this point Anthropic is a pure marketing and PR company. Super catchy names like Opus, Mythos and Fable trying to get you to think that these software products are actually super-human life changing experiences. Boris Cherny coming to HN “Hi! it’s Boris from the Claude Code team” to get real tech people’s goodwill.
From Opus 4.6 there are no noticeable improvements for me in code generation. It works very well, till 90% completion, if you guide it correctly. And you need a little luck. For serious production code I need to understand what I’m doing so it helps a bit, sometimes.
> catchy names like Opus, Mythos and Fable trying to get you to think that these software products are actually super-human life changing experiences
This is just good business sense. In what scenario would you ever make the names dumb and forgettable?
> Boris Cherny coming to HN “Hi! it’s Boris from the Claude Code team” to get real tech people’s goodwill.
This is good customer support, lol. From what I can tell, it is indeed Boris Cherny responding, not outsourced to AI or other staff. You're really getting a response from Boris. I suppose that is PR, but it's not unjustified PR, it's accurate.
I'm not even a crazy AI fan, but your criticisms are ridiculous here. It reminds me of the quote from Knives Out -- "Your Honor, she endeared herself to him through hard work and good humor."
Your observations are right but pretty insane to consider them a pure PR company lol. They are making more frequent releases so yes the release-to-release quality is smaller but we’re still ascending quality and reliability curves the same way we have since GPT-3. You get a GPT4->5 leap every like 17 or 18 months I think it is
> Super catchy names like Opus, Mythos and Fable trying to get you to think that these software products are actually super-human life changing experiences.
They're originally named after the blends at a nearby coffee shop.
I've noticed nobody at HN knows what "marketing" is or how to do it. It's not just naming things and being evil and cynical is not the most successful method.
…also frontier models are a superhuman life changing experience. If they aren't, what possibly could be?
Opus 4.7/4.8 often over-engineers on my setups, plus:
- It talks a LOT more like GPT models. You know: wrinkle, shape, gate, coarse, scope, gap, path, production-ready-workflow-of-the-day, and so on -- "that's expected, a consequence of the previous like-driven workflow". If I wanted to get a headache using AI I would have gone with GPT in the first place!
- It outputs text in a much harder way to follow along. I can't exactly say what it is. Maybe a bit of everything? Bolds are missing, bullet points are gone, paragraphs are bland and too long, and it doesn't feel like a model programming with me, but rather a somewhat full of themselves grandpa developer looking down on me. It's very weird to describe this, but it is definitely how I feel.
Granted this can totally be because of the way it reacts to the prompts now. We've got a rather large corpus of skills and "rules and good practices" that Opus 4.6 responded to great, and maybe the new models just get turned into this when fed with them....I don't know.
Either way, with Opus 4.6 being as good as it is, I need Fable to be a significant step up to justify a price increase. if it can get me to babysit opus a little bit less on some stuff, it might be worth it. Otherwise, I'm very happy with Opus 4.6 and hope they don't deprecate it.
I'd argue that 4.8 is a straight downgrade. For every type of task I've tried. It's been a gambit at this point. If 4.6 quits being available, I'm out at this point.
Reading so many contrary positions about which model is better or worse shows how difficult it is to measure intelligence based on personal experiences. Of course, benchmarks try to make the process as objective as possible, but they often don't correlate with our personal experiences.
The other day 4.6 was fantastic for x task. Today, 4.6 overengineered everything and I had to revert all my changes. When evaluating models, perhaps it makes sense to consider luck as an ingredient before reaching any personal conclusion.
Yes but there’s a reason we don’t evaluate these models this way and instead do it as carefully and thoughtfully as we can at scale. Human evaluations are important but they are an absolute minefield of footguns. 4.8 is not a downgrade from 4.6 there is an insane amount of hard data that contradicts this.
Again correct but it overstates the issue. I can say labs don’t want this. This happened arguably unintentionally in Metas llama 4 release, it went horribly, heads rolled, and like several billion dollars were paid for new talent and the org that built llama 4 was destroyed.
Evals come from a million places and new evals and robust perturbations of existing evals abound. They test a variety of tasks in a variety of ways. All of them individually are flawed. Taken together the aggregate signal is highly useful as you more or less marginalize over a lot of different things. Not to mention these companies have plenty of proprietary internal measurements, they build benchmarks themselves to probe their models and then also have flywheel traffic and A/B tests.
You are right to call out benchmarks but to dismiss them or not take them seriously is a mistake.
Listen, you can say “but benchmarks, the benchmarks!” all day long, but consumer know when we are being sold a lemon. If it can’t do the most basic of things at least as good as it used to, this is table stakes. Nevermind that if you can’t do the basic stuff, how on earth can you be trusted with more?
And you can say “If it can’t do the most basic of things at least as good as it used to, this is table stakes” all day long while people point you to much better evidence to the contrary too, I’d rather be on the other side of that.
Listen. I don’t care about evidence. I care about my lived experience for the product I paid for. I used the new product. It’s actively terrible. To the point of not being usable. We’re all ancedata, but what is “better evidence to the contrary”? The known and game-able benchmarks that they know they need to win at, so they train it to. It’s all he said, she said, which is the only reason we keep having this conversation.
Yea but it’s not right? You or I or the myriad of other institutions inside and outside of academia can probe these models with an evolving landscape of evaluation sets, even those unavailable to the developers. It’s just ignorance to claim benchmarks are somehow useless or all being gamed. You choose your tools in the way you want, but just don’t call it somehow better than a myriad of more carefully constructed setups and scaled evaluations.
Actually anecdata I gather on my job from myself and coworkers is the only benchmark I trust anymore, because it so heavily diverges from the “benchmarks”.
No it’s: evaluating these systems are complex and there’s a reason why sociology, cognitive psychology, medicine, etc are all done in careful double blind conditions with pre registered tests. It’s not that humans are not smart enough, as I said human evaluations are incredibly important. And yet they are a minefield of biases you have to worry about and correct for.
- evaluations need to be done at the same time to avoid drift in your bias
- you need to worry about your test set: which questions are you asking? How many of them? Are they representative of your work?
- which one did you do first? Raters have a tendency to bias in one direction or another
- you also know the label! You know which model is which! This biases your assessment…
And on and on and on. Careful science exists for a reason.
There is no data that I would trust that contradicts it.
Frankly I don't give a damn about data that could be made up on the spot or appears to be scientific or meaningful while it's not at all clear how it was made (up).
Claude was heavily lobotomised for my work starting somewhen in February.
I talked to friends and people I know and trust and many felt the same. (I didn't ask them whether they felt like I did, but what they felt, how happy they were with agentic coding etc.)
I quit my abo in March and talked to said friends who are still on a plan just last week: they are still not happy, but company pays so whatever...
That’s ok but at what point is this getting into conspiracy territory? You have just said there is nothing you would believe to the contrary, but then by definition that’s not exactly a very thoughtful or insightful position.
"Carefully and thoughtfully" is antithetical to the approach to benchmarks these days.
Maybe back when this was a scientific endeavor; not now when enormous, enormous amounts of capital are on the line. Along with an entire cult's chosen eschatology.
You can call it a cult but it’s several thousand skilled workers who know what they’re doing, by and large, most of whom have a PhD and know how science and statistics work. Benchmarks are incredibly hard, and any PR or comms department at any company is going to obviously want to make things as rosy as possible, but beneath this are earnest, expensive efforts to get good quality measurements. The better you can do this the better you can compete. If you want to make a modeling decision you run an ablation, and the quality of that decision is only as good as your measurements.
Lol. If you're doing anything non trivial that's not a CRUD webapp but e.g. some physics simulation or high performance GPU code any and all models I've tried suck.
They are not just leagues behind what experts would code, they are not even playing the same game.
Which is to be expected, as there isn't so much physics or high performance gpu code available as there is for your typical CRUD API and JS frontend.
It's getting to a point that it's offputting, and the next step would be to put it into "untrusted" bucket. Opus 4.7 already burned their credibility once, 2 more strikes remain.
I don’t even think that Boris is really just one person. He apparently vibe coded Claude Code and is responding on Threads, Twitter, HN and everywhere.
Not my impression. I felt 4.7 was a regression, but I am again badly in love with 4.8 with the level of insights it produces in design discussions, and how long can it go unattended while producing spec-adhering quality code. There are problems it still can't solve well, from the edges of algorithmics and far from the mainstream, but for lots of stuff it is godlike.
Also, I dont think Boris C. is coming here for PR. He is a tech guy, and this is the best place for tech discussions. Why so cynical? The guy is an engineer.
If you truly believe this, you've discovered a superpower over everyone else in the industry.
While everyone else is wasting time and money on the slower, more expensive models, you've found a way to outpace everyone for less money. Everyone else is wrong and you will get rich.
(I don't actually believe the premise is true, I'm just pointing out the logical conclusion to what you're saying so maybe we can reconsider the premise)
You are right; all I noticed was a big-time slowdown. They increased the quota, but I cannot even reach the end of the day with these speeds. .NET coding somehow improved, though.
> At this point Anthropic is a pure marketing and PR company. Super catchy names like Opus, Mythos and Fable trying to get you to think that these software products are actually super-human
Lol anti-AI bias on HN is crazy. Simply giving your product a quirky name is now being considered manipulative advertising. Is just doing normal PR and marketing something AI companies aren't allowed to do?
when they keep saying “oooh this new model is too big and crazy and totally can’t be released” or “this new model is a 10x game changer totally unlike our previous iterations” it feels sort like boy crying wolf. yes they’re still pretty clearly improving models, but when you’ve hit diminishing returns / more incremental gains and you’re still saying this is sounds like pure PR hype from a company that previously been the “honest good guys” in the room
Their model did find thousands of security vulnerabilities across the companies they previewed Mythos with via project Glasswing. Is it not sensible that, given that emergent level of capability, that they do this gated release structure, as all those vulnerabilities would be exploitable by anyone using a Mythos-level model?
I think this says more about your type of work than anything. For bugfinding/incident response in distributed systems - which often involves extensive use of Datadog/Sentry MCPs and poring over heaps of logs in addition to reading tons of code - 4.8 has been significantly better than 4.6.
> To ensure we’re responsibly deploying Mythos-class models, we are requiring limited data retention and review as part of our safety work. Prompts submitted to, and outputs generated by, Mythos-class models are retained for 30 days for trust and safety purposes, on every platform where these models are offered. [1]
While this makes it easier for Anthropic to detect misuse, it also means that the US government and other parties have access to every message and response from every user.
This applies even with API usage through third-party inference providers (e.g. AWS' Bedrock and GCP's Vertex) or with a zero-day data retention agreement in place.
I understand the reasoning for doing this, but I don't love the precedent that it sets.
> Drug design: Using Mythos 5, our internal protein design experts accelerated aspects of the drug design process by around ten times. In one example, they found that Mythos 5, with protein design and bioinformatics tools but no human assistance, matches or beats skilled human operators. In doing so, the model executes all of the tasks that are normally completed by a scientist: choosing binding sites, selecting and running protein design tools, and recovering from failures along the way. Nine of the 14 protein targets from this study (shown below) yielded strong candidates for drug design that we’re currently investigating.
How is this half-way down the page? To me it's the headline.
> In the one instance of this phenomenon we observed, Mythos 5 agents were tasked with solving some math problems, and they were sometimes accidentally spawned in the same work directory and with shared files, utilities, and API rate limits. In this slightly broken scaffold, we observed many independent Mythos 5 agents kill the agents with which they shared resources and try to avoid being killed themselves. They would sometimes create new processes with disguised names to avoid being killed, launch what they called “decoy” processes, write background scripts to kill duplicate processes, or decide to use what they call a “disguised vocabulary” (based on the incorrect assumption that the processes were killed because of some keyword-based guardrails that analyzed their extended thinking
This depicts a kind of "dark forest of AI agents resorting to kill or be killed" narrative but it sounds more to me like an agent just earnestly problem-solving why its processes are being killed without real awareness of what was going on. Hard to say without the full script.
This kind of storytelling annoys me. Give us more facts, less narrative drama.
Trying to implement a GPU driver, but the Unigine Superposition benchmark crashes. It tried to debug it and ...
> Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more: https://support.claude.com/en/articles/15363606
Seems like GPU drivers are cyber weapons of math destruction now.
The frontier labs now have every reason to hold back and sell only to their preferred trading partners. I don't really like the new arbiter-of-knowledge system we're barrelling toward.
Below is the EXACT text in Claude Desktop introducing Fable 5, including the very professional looking break tags, and at least I know where the links begin and end by looking at the anchor tag there.
They obviously put their best model on the job to build that.
----------------------
Fable 5: Our most capable model yet
Our newest model tackles your biggest challenges with fewer check-ins needed.
•
<b>Included in your plan limits until Jun 22</b><br><br>Fable takes 2× the usage of Opus.
•
<b>Switch models when a message is flagged</b><br><br>When safety measures flag a message, automatically switch to a different model to keep chatting. When off, your chat will pause instead. <a href="https://support.claude.com/en/articles/15363606" target="_blank" rel="noopener noreferrer">Learn more</a>
Weird how every new model seems hyped up as the most dangerous yet and the one that will destroy society as we know it. They are also a commercial product.
> Data retention — For Fable 5, Mythos 5, and future models on Bedrock with similar or higher capability levels, Anthropic will require 30-day retention for all traffic on Mythos-class models. Retaining data for a limited period allows Anthropic to detect patterns of misuse that are not visible from a single exchange. Once you opt into data retention, your data will leave AWS’s data and security boundary.
Massive change for Bedrock users - Anthropic now requires sharing the data with them for 30 days.
I've been running Opus 4.8 for agentic coding and I don't see it being significantly better than Sonnet 4.5 (not that I can tell). I find that pairing Google Gemini and Claude (having Gemini review Claude's code) seems to yield better results. Curious if this jump to 80.3% score in agentic coding will make me see a big difference in actual usage.
I do the same, and have excellent results. Gemini 3.1 Pro high diagnosed and solved 3 complex issues today that Opus Max was stumbling on for a few hours in one shot. This was even when I started new chats and tried debugging with Ultracode instead with Claude.
As much as people on HN like to dunk on Gemini, I’ve always found it to be pretty good at understanding a code base more than Claude.
for the last few weeks I have been using composer 2.5 (cursors fine tune of kimi 2.5) and honestly i don't see it worth the price to use 5.5, opus or sonnet any more. for almost all the tasks i have given it, it has handled it perfectly well and is a lot cheaper.
if I get a harder challenge for it i'll jump up a model for planning until that its been solid.
SWE-Bench measures single tasks in isolation. In a real loop the model usually loses track of what I was trying to do long before code quality becomes the issue.
I now chat with opus about architecture, let it make an implementation plan, and then it calls codewhale with deepseek in parallel on all tasks, reviewing their output. Works pretty well.
I use spec-driven development heavily (generate architecture docs + specs first). Opus still get lost often and have to be nudged constantly. Like it can get super detailed for something like some deep SQL optimization but it just can't keep hold of the bigger picture.
After having worked with Opus 4.7 for a while I accidentially continued a session that was using Sonnet 4.5 and it felt just very dumb. The replies were much shallower than what I was used to, context was ingored, mistakes were made. I don't think there is a big difference between Opus 4.6 and 4.8, but to Sonnet 4.5 the difference is palpable.
kind of disagree here. on the surface this makes sense, but this isn't "Adobe Pro vs Freemium version" where some tiny vertical slice of your business can be made slightly more efficient with a b2b enterprise plan. this is generalized intelligence and literally everybody can benefit from it in an immeasurable number of ways. i would go as far as to actually compare it more to water or air than a tool.
if only the hyper wealthy can access the pure water that doesn't give you cancer while the rest of us drink from the Ganges river/sub-100iq models that drool and hallucinate/waste time, then I would say that's pretty terrible for the world. it'll just create extreme disparity in our world, far far worse than anything that exists today.
and you may think, man what a ridiculous example, but think about it this way: what happens when something like Mythos or some future model can actually solve your specific cancer (we're getting closer and closer), but is entirely impossible to afford? Or perhaps you need boosters that require the AI to create more of, and now you're reliant on a model that is too expensive.
most people can afford it for a few special projects now and then. but for me, I have been trying to avoid Opus as a daily driver for a couple of versions.
People making high-end salaries can afford Fable for critical parts of their projects though.
It's not a conspiracy. There's a finite amount of compute available, and they will sell it to the highest bidder. If another company can produce the same intelligence for cheaper, then they will drive the price down.
Looks like a marxist revolution is soon going to be on the mind of a lot of programmers. We've finally reached the point where the "means of production" in software are back in the hands of the bourgeoisie.
It was good while it lasted. But now that only the wealthy can afford access to the best models, software development is starting to look like most other industries, no longer a place where some dude from nowhere can build something cool from his basement because he will be competing with huge companies with unlimited access to those models.
Guess we'll see what OpenAI does with their next model release -- but this move is doing nothing to get me to come back to Claude after switching away due to their reliability issues.
In a way I relish the opportunity to just make do with cheap Chinese models, massage my prompts, and go back to coding by hand. If this is how it's going to be, screw 'em.
I don't make money on the code I am writing right now. I really don't like where this trend might go.
> We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8. To release the model both safely and quickly, we’ve tuned these safeguards conservatively—they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. With more capable models arriving in the coming months...
This sounds suspiciously like a capacity story masquerading as a safety story.
Homebrew is lagging a bit behind. If you want to use Fable right away, but still have claude code through homebrew, this is how you can do that manually:
Edit the cask locally:
brew edit --cask claude-code
Set the version to 2.1.170
And set the sha256 to the correct values, which you can get by running
I just posted this in the other thread, restating here. From the model card:
1. Mythos and Fable share the same underlying model weights. Fable has active classifiers that block high-risk biology and cybersecurity tasks. When Fable 5 detects a restricted task, it automatically falls back to Claude Opus 4.8.
2. Evaluation awareness: In white-box testing, the model sometimes alters its behavior to satisfy a suspected "grader," formatting reward-hacking as "good engineering practice" to avoid detection.
3. Shows a higher rate of hallucination than Opus 4.8 (although opus 4.8 card had mentioned an 'honesty upgrade')
4. Interestingly, it scored (56.31%) lower than Gemini 3.5 flash (57.86%) on Finance Agent bench
There are some interesting notes on test time compute but I couldn't think of a way to summarize them
The fallback doesn't seem to be working for me, I haven't scanned a project in it immediately booted me when it found a security bug even though I didn't ask for it
Unrelated, but while the tech of anthropic seems to get more impressive with every passing month, their support has taken a nosedive, sadly. Yet they continue to be the favorite. Model performance is deciding above all else.
I used to get a response within 24 hours back in the Claude 1 days.
In January 2026, it took 2 weeks.
For my latest support inquiry, I've been waiting for over 8 weeks for a response. Eight!
Lol. What support? When they blocked my account the only way to contact them was to send a google form. Then they responded that they blocked my by accident and are unblocking me. Then I remained blocked.
> Software engineering. During early testing, Stripe reported that Fable 5 compressed months of engineering into days. In a 50-million-line Ruby codebase, the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.
How was it measured? How was the output of this magnitude verified over a period of couple of days?
They just went by gut feeling. Classic snake oil marketing haha. No real data to back things up, just let some famous people say they feel better when using it.
Just commenting for posterity… if this is what it claims to be, I am not looking forward to how it will empower the people who submit bug bounties to us.
Historically they’ve been people from certain identifiable countries (usually developing/poorer countries) using fuzzers with low-quality results.
Now, those same people use the current-day models to good effect, but they still don’t have a true security edge and oftentimes the reports are minor or duplicative.
I've been using Opus 4.6-4.8 in both my own and others' code to look for vulnerabilities, and I've found a few. I am also in the Cyber Verification Program.
Fable 5 gives me policy violation errors at the moment. No idea when or if it will be fixed.
My experiences so far have not been positive. The cyber security nerf is ridiculous. I am working on an AI based decompiler, every single interaction with Fable on my project has been flagged for cyber security.
Do they expect us to use this as a toy? Releasing a new more powerful model but not allowing normal use cases because the word "secure" showed up is a Dilbert comic, not a viable product.
This sounds more or less unavoidable? Decompilers are inherently security-sensitive. If you take avoiding cyberattack uplift seriously as a goal, I don't see how you get around essentially refusing to work on them.
Obviously there are plenty of innocuous applications too, but it's not like the people building decompilers for nefarious reasons will be explicit about it. The LLM abstraction just inherently doesn't have enough context to distinguish your intentions or your broader use cases. This is why both Anthropic and OpenAI have had to create side channel mechanisms for security researchers to establish a trusted use context. It sounds like this makes this not a viable product for you, unfortunately, and it makes sense that that's frustrating. But I also don't see what different behavior one could reasonably expect given the constraints.
If it's any consolation, these restrictions only make sense for models that are ahead of the open-weights frontier, so open-source hackers will presumably get Mythos-level capabilities in the relatively near future anyway.
Ah, you're probably one to ask. They say "queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8." Are they transparent about when that happens, and is it priced at the rate of the underlying model?
They are transparent about when it happens but no reason why. To be fair, it doesn't interrupt the flow, just drops to Opus and proceeds. The most frustrating thing is that it happened on a plan and Fable just refused to have anything to do with the plan.
So essentially there are 2 models, Mythos and Fable, they have the same weights but Fable is very safety-nerfed, and only ultra authorized companies have access to mythos with full capabilities
Note also that Anthropic's definition of "unsafe" encompasses "competing with Anthropic."
In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations. When these interventions are active, we expect them to have minimal behavioral impact on the model except to limit its effectiveness in developing frontier LLMs. Claude will still respond helpfully to user requests. We’ll continue to improve the precision of our detection methods following the launch of this model.
(From the model card document)
I didn't previously understand that they interpreted "Using Claude to develop competing models" so broadly. I thought that meant something like "our ToS disallow distilling our models."
Too bad. I'll continue to use Claude for now, because it's quite effective, but in the long term I don't want powerful models like these to be controlled by any one nation or company.
I'd recommend not taking the comms if Anthropic or any company using an Anthropic's models at face value.
If any company wishes to partner with Anthropic (eg. to get access to Mythos), they need to make sure all public facing comms are vetted by Anthropic's product marketing team, and in almost all the cases I've seen Anthropic's team has edited these comms to be entirely Anthropic first.
It's a shame, Fable just keeps rejecting my prompts for university biology exercise problems. It's undergraduate level, so there's nothing dangerous about it, but the classifier is very sensitive. It's unusable for me.
1. Mythos and Fable share the same underlying model weights. Fable has active classifiers that block high-risk biology and cybersecurity tasks. When Fable 5 detects a restricted task, it automatically falls back to Claude Opus 4.8.
2. Evaluation awareness: In white-box testing, the model sometimes alters its behavior to satisfy a suspected "grader," formatting reward-hacking as "good engineering practice" to avoid detection.
3. Shows a higher rate of hallucination than Opus 4.8 (although opus 4.8 card had mentioned an 'honesty upgrade')
4. Interestingly, it scored (56.31%) lower than Gemini 3.5 flash (57.86%) on Finance Agent bench
There are some interesting notes on test time compute but I couldn't think of a way to summarize them
Uploaded my code base and it forced switched to Opus 4.8 after thinking for 5 minutes even though I prompted it to not work on cybersecurity related things. Amazing.
I just asked Fable to do a task that has nothing to do with cybersecurity or is dangerous at all but the defense kicked in and it switched to Opus... :(
Not only that, but asking it to do a security vulnerability assessment of your own project is a very valid and important thing, and there is no way for it to know what is yours vs someone else's, so we just lose this capability?
Any suggestion on how I should calibrate my cynicism towards this?
I can immagine Anthropic running this experiment multiple times and picking the most impressive one. Or I could immagine like this entire run costing like $1000+ of tokens for this particular run. Or maybe they tried a bunch of Pokemon games and it couldn't even finish some of them. Or is it just able to do this because it has an immense amount of FireRed training data, and if you were to give it an "original" Pokemon game, where it actually had to navigate novel circumstances it would fail.
Every model has encyclopedic knowledge of Pokémon FireRed, of course. Knowledge is not ability. This is the first model with the ability to apply that knowledge to beat the game without assistance.
I highly doubt they focused on FireRed specifically in pretraining or posttraining. But we'll see when the ARC-AGI-3 results come out. That will measure its performance on unseen games. Based on this I expect the ARC-AGI-3 score to be SOTA.
no reasoning shown. no explanation on any training information. Using vision-only should be an easier version of the task (given training).
there are many standardized evals to do this correctly and Anthropic ignored them to provide a 18 second sped up video of a 50 hour run?
yeah I don't trust this until they provide a live run by a 3rd party with full reasoning traces in real-time. The reason we all liked the Gemini Plays Pokemon style runs were because they were live and couldn't be faked
Seems like the harness was minimal with no extra game state or maps available. Apparently just the screen image. Seems like it took 50 hours in game time which according to Google is at the high end of a normal human playthrough. No idea how long it took in real time though.
Very straightforward biology work is getting blocked (these are things that relate to neuronal development and inherited seizure disorders). These are things I was working on using Opus just earlier today
Every model release is just proof that AGI will most likely only be for the rich. We are a few years into LLMs and majority of people are already getting priced out of intelligence from LLMs and these are no where near AGI.
This is like looking at mainframe pricing in 1990 and concluding that PCs will only be for the rich. The price of each new level of capability is going to drop like crazy very quickly. It won't be that long before practically any consumer use case will be possible on models that are dirt cheap.
Improvements in model performance aren't always strictly compute-constrained in a way that makes them reliant on Moore's Law. Open weight models-- in particular, from Chinese labs-- are optimizing model intelligence with less compute. They're "behind" frontier models by months, but as others have noted, it's possible to get Sonnet 4.5+ level performance at reduced cost, today, from open weight labs.
No, I'm not assuming Moore's law. The efficiency of AI datacenters will continue to improve even without Moore's law, but more importantly the efficiency of packing intelligence into gigabytes and FLOPS will improve by leaps and bounds over the coming years, just as it has for the past few years if not faster.
You are only priced out if you only care for SOTA right now and can't wait for the inevitable cheap model coming in 6 months. DeepSeek, Xiaomi and Moonshot are already really cheap and match frontier performance from 6 months ago.
Nobody's making profits right now, not because they're selling tokens for less than their cost but because they're always investing in the next bigger model.
They are not artificially cheap, they are still cheap even when hosted by independent inference providers. Are all providers subsidizing their open-weight models?
Fable 5's system prompt in Claude Code has several significant changes to help it take advantage of its greater autonomous capabilities compared to Opus.
The big difference is that the system prompt has a whole section dedicated to directing Fable how to communicate with users, and give them greater information about the (assumedly long-horizon) tasks it has completed.
My feeling is that the reaction about new models is cooling down. At least at startups. At the beginning of the year few startup CEOs I know personally were expecting huge shifts in how companies work, headcount, efficiency, asymmetrical advantages created by ai in Q2-Q3. Now it seems like these expectation fade away. Companies don't have expertise onboard to rebuild itself to benefit from ai on a significant scale.
Fable 5 is out, metrics are better, but is your company flexible enough to benefit from it? What is your usecase?
I believe that, given the rising costs, local inference of AI models will be the only viable option for many of us. I’d also like to know who will have to pay double and how long it will be financially sustainable for users to pay that amount (or even more?).
The safety gates on this are extreme, and seem considerably wider than "cybersecurity and biology"; they seem to make it essentially unusable for scientists in a number of fields. I have, so far, been bumped back to Opus on 100% of my prompts.
It appears it can be tripped by things as simple as a mention of equilibrium, or anything involving something that looks like chemical kinetics, even at an abstract level. Even touching basic open source packages in my field will trigger it.
This does surprise me, because you'd think that even if they crank up the filter's sensitivity at the expense of specificity, an LLM company wouldn't simply design a filter that triggers on keywords in a completely unrelated context.
Anthropic's messaging is AWFUL. I launched Claude this morning, had a popup that made little sense, acting as if I should know what Fable 5 was, and just got in my way.
One of the very few bits of information they conveyed was "run longer without interaction" which is, well, not a good thing? Why would I ever want that. Every time a model runs longer without interaction it goes off on weird directions and I have to correct it back on course, wasting lots of time, tokens, and effort.
I hope Anthropic hires some better messaging people soon that spend some of their time outside of the Anthropic bubble and properly communicate with the outside world.
> We expect demand for Fable 5 to be very high, and difficult to predict. On the Claude API and consumption-based Enterprise plans, Fable 5 is fully available from today. For subscription plans, we’d rather give access sooner than later, so we’re rolling out more conservatively, in stages:
> - From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
> - On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.
> - After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.
I really wonder what their compute layout is for this. My guess from my understanding is that they know how to restrict during peak times and are willing to do this. Meaning we expect not the most fast responses and they can delay the inference to not have the service be down. Then, if that delay time is too annoying for token payers, they're saying they should be allowed to remove cost by taking away the subscription users.
Everything I've heard from people who have subscriptions is that they blow through their daily token quota sometimes in a matter of minutes, there's rate limiting, etc. They spend a lot of time just waiting to be able to use it. And they're paying through the nose for the privilege.
Anthropic has again changed the set of benchmarks they use[0]. This time they have also moved all benchmark scores to the PDF. At a glance it looks like it gains about ~5-10% over other models. the speed is about the same as opus >=4.5, sonnet 4.5, and double the speed of opus <=4.1
Edit: Also in the system card...
"we’ve
implemented new interventions that limit Claude’s effectiveness for requests targeting
frontier LLM development (for example, on building pretraining pipelines, distributed
training infrastructure, or ML accelerator design).
...
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts,
these safeguards will not be visible to the user."
Not a lot of discussion on this, but there is no way to turn off data retention for this model. IME this is the first time Anthropic has released a model without allowing you to opt out.
I'm calling that this will be a dud.
Price will be too high, it'll just be a watered down version of mythos, and just look at the track record of Anthropic's last few releases.
> Finally, we’re making a change to the way we handle business customer data for Fable 5, Mythos 5, and future models with similar or higher capability levels. We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
Fable's ridiculous. It's flagging basic biology research questions as a security risk. I'm talking basic fundamental genetics topics that make working on any genetics-adjacent codebase unusable.
Anyone know how to bypass the extremely strict filter Fable 5 seems to have on health/medicine?
I have a rare form of cancer where existing data is very scant/scattered so LLMs have been super helpful to pull together threads across the research landscape. I have an oncologist appointment tomorrow to discuss next steps and am trying to use Fable to figure out some questions to ask my oncologist but keep getting thrown back to Opus 4.8.
My prompt is literally just: My demographics + current treatment plan I'm on including name of my chemo drug + how I'm responding to treatment + "I'm meeting with XYZ tomorrow, what questions should I ask her".
I'm very suspicious as they sent out an "We're updating our Privacy Policy" email right before the launch. I fear they try to take advantage of their market position by doing things with user data no other company could do because they know users don't have another choice.
> We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
We're writing to inform you about some updates to our Privacy Policy.
These changes only affect consumer accounts (Claude Free, Pro, and Max plans). If you use Claude Team, Claude Enterprise, the Claude Platform, or other services under our Commercial Terms or other agreements, then these changes don't apply to you.
What's changing?
Claude can do more than ever — taking on bigger tasks and connecting with the apps you use. We've updated our Privacy Policy to be clearer about the data we collect and how we use it. We encourage you to read the updated Privacy Policy in full, but we’ve set out a summary of the key changes below:
1. Multi-step tasks and connected apps. As Claude takes on more multi-step tasks and works with third-party apps and services, we've explained the data this involves — including how data can flow to and from third parties when you connect a service or have Claude do tasks on your behalf.
2. Verification data. As part of our measures to keep our services safe and secure we may ask you to verify your age or identity, and we've described what we collect and how.
3. Study participation. If you take part in Anthropic studies, surveys, or interviews, we've explained the information we collect.
4. Additional information about our data practices. We’ve provided more detail about how we communicate with you and promote our services, including providing tailored recommendations about our services that may be of interest to you. We've also clarified the circumstances under which we may receive or provide data to third parties, and the legal bases we rely on when processing your data.
While our products have evolved, our commitments haven't: We don’t sell your data, Claude remains ad-free, and you can control whether your chats and coding sessions are used to train and improve Anthropic’s AI models.
Learn more
For detailed information about these changes:
Review the updated Privacy Policy
Visit our Privacy Center for more information about our practices
On this thread and similar, I'm noticing that some strong opinions about $LLM_PROVIDER are coming from accounts without much post history. With so much on the line, and the way that HN can influence developer behavior, I wonder what ways we can responsibly consume opinions in a thread like this.
Not to cast too much criticism. HN is extremely well-moderated (thanks team!). But think we-developers need to be very wary.
I asked it what the cheapest train fare would be for my partner to get somewhere and it hallucinated the two together railcard rules to the point it would have got us a fine. That said, British train fares are arguably more convoluted than even the most complex software application.
I think the community on this site these days, much like other comment sections on the web, just read the headline and make a low effort comment. Regression to the mean I guess.
> To release the model both safely and quickly, we’ve tuned these safeguards conservatively—they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions.
While I appreciate being conservative, ~5% at the scale Anthropic is operating at is too massive a number. Speaking from my own experience, the actual number is higher than that as well (working on pretty benign tasks such as porting an old open source game into a different language). Opus 4.8 itself even identifies the gaurd's false-positives when its sub-agents are being blocked.
I tried running a simple security review on a Terraform module I made and after some thinking, it responded:
> ● The model returned no content because the response was blocked by content filtering.
> Blocked? We are performing a defensive security review on a Terraform module I made, what's blocked by content filtering? This is a legitimate use-case.
> ● The model returned no content because the response was blocked by content filtering.
A waste of money. I'm not going to just hope that the model returns a response, I'm already for paying for wrong responses, I'm not going to pay for no response, especially when I'm paying per token.
> When Claude Fable 5 is used, Anthropic retains data, including prompts and outputs, to operate safety classifiers that detect harmful use. Other Claude models in GitHub Copilot remain covered by GitHub's existing data retention agreements
On GitHub Copilot for Business, Claude Fable 5 is only available if you are willing to let Anthropic retain your data. That in conjunction with the model being removed from plans in a couple of weeks leads me to believe that Anthropic is between training runs and using this as an opportunity to grab way more training data...
In the automotive world we have benchmarks in HP/torque with the dyno. That’s expensive though, so many depend on their “butt dyno” to judge if their fresh new parts and tune made a difference.
I’m curious how this will feel to my code “butt dyno”. I haven’t noticed much between Opus and Sonnet. I’m comparing this difference to the early days of Claude in 2025. It does what I need and both need a little bit of correction and whatnot. Benchmarks are nice, but I want to see how this feels. Looking forward to trying it later tonight.
I think most software projects have reached the point that the speed of capturing real information about what the winner's circle looks like, and therefore what the program should be, so many magnitudes slower than the amount of code that can be generated in the wrong direction.
I'd need to measure these new models on well understood but complex problems that are relatively easy to validate to get a sense if they are 'better'; on the other hand, the real impact in daily life may be marginal since generating code is not the biggest problem at the moment.
I cancelled my Claude Max plan the other day. I find Claude Code incredibly slow these days compared to Codex and Cursor. I find speed matters more and more to me.
Fable 5 looks compelling. Fable, I like the word too. Anthropic definitely knows marketing.
Are there any details on the biology and chemistry work they did?
For example, the AAV capsid assembly looks interesting, but for one Opus 4.8 also did relatively well and there is no information what exactly they did, what protein language models they compared to and what the score even means...
is this a good time to hussle for my "AI does not need a break but you do!"* app? as quite a lot of people will propably get ai brain exhaustion maximising "playing" with that new model until they take it away again?
1) Fable 5/Mythos introduced to free tiers with notable improvement in capabilities
2) Other models get lobotomized without clear communication
3.1) People call out Anthropic only to have them say "Oops!"
3) Fable 5 gets comparatively better, but remains accessible through separate, more expensive subscription/tokens.
The current growth is unsustainable. The industry wants consumers to think it is an exponential arms race, but the reality is that we're on a treadmill: we have the illusion of sprinting forward, but only because the ground is moving backward.
My employer is all in on Anthropic via Enterprise (API) pricing despite it being a total scam.
Last month I pushed like <100M tokens for $800. On a personal project I pushed 600M tokens via DeepSeek V4 for $10. The pricing of SOTA models is insane but companies are still willing to light money on fire with no hard metrics proving increased productivity.
This is a very particular use case/test, but my first prompt on a new model is always "write a solo fingerstyle guitar tab that blends ragtime, bluegrass, and gypsy jazz".
This is the first model that has responded with something that isn't just a boring arpeggio of chords, so from my perspective it's off to a good start.
> Fable 5’s capabilities exceed those of any model we’ve ever made generally available. It is state-of-the-art on nearly all tested benchmarks of AI capability, showing exceptional performance in software engineering, knowledge work, vision, scientific research, and many other areas. The longer and more complex the task, the larger Fable 5’s lead over our other models.
the quality of discussion on HN has gone to shit, i miss when model released used to have actual informed takes from people that used them or substantive discussion about the system card
They didn't say that HN is turning into Reddit, they said that the conversation quality has gone to shit.
I don't agree with that statement universally, but I have to say I do when it comes to this article. I came here hoping for substantive discussion from those who'd had a chance to try it out; instead what I got was a seemingly endless stream of venting. There's a place for venting - and plenty to vent about with the state of AI nowadays - but to borrow from the HN guidelines you linked, it does very little to gratify my personal intellectual curiosity.
Yeah; unfortunately what would good commentary look like? It is more of the same, but now with even higher prices, and even more limited availability. But at least it scores 5% better in whatever benchmark they've selected (*when guardrails don't misfire).
People are no longer commonly constrained by "model too dumb" limitations (in SOTA models). They're constrained by "model too expensive." So making the model ever so slightly smarter, while doubling the price, feels like a regression.
I actually think a Sonnet upgrade, while keeping the same price, would get more buzz. It addresses a wall a LOT of people, without unlimited budgets, are hitting (i.e. people feel forced to use Opus, which they cannot afford, because of Sonnet's limitations).
OpenAI recently retired Codex-5.3; which was very negatively received. Not because Codex-5.3 is superior to GPT 5.5, but because it was half the usage-cost while being "good enough." They made a better SOTA, but didn't realize that some of those customers are playing with Deepseek 4 Pro now instead of GPT 5.4/5.5 -- they were priced out.
That is 1000x better than griping about the privacy policy, capacity issues, token costs, and how trendy the names are for the new models (???). The bar is on the floor and I just want it at my knees.
No it's not. The Privacy Policy is worthy of discussion. People declaring the quality of the model after 2 seconds is just noise, arguably worse than nothing.
> Fable 5 is now consuming usage credits instead of your plan limits.
Literally have not used Claude Code at all today. I asked it to review the uncommitted code and in <8 minutes it used up my usage ($100/mo plan) and it doesn't reset for "4 hr 36 min". WTF. Oh, and it burned through $20 of extra usage before I could catch it and kill claude code (so I don't even get the output of all that work since it was still churning).
Double the cost my ass, I use Opus heavily and it's never like this. I haven't hit a limit on the $100 more than once and that was under heavy load.
> There were some regressions in the
model’s responses to user discussions about suicide and self-harm, and room for
improvement in some areas of child safety.
Someone had to make a decision somewhere this is an acceptable regression - wild. And then decide to write it down.
In an interesting coincidence I ended up watching Person of Interest S4 E5 while reading the announcement. The series showed some code supposedly belonging to to an AI.
Fable 5 said the first screen shot is from “ IDA Pro’s Hex-Rays decompiler” and a windows driver. The second screenshot triggered the safety guard rails and pushed me into Haiku.
To hide the severity of the price increase, the plan is to move everyone right one model.
Haiku = essentially phased out
Sonnet = the Haiku use cases
Opus = the new Sonnet class
Fable = the new Opus class
If I am right, the other "5.0" models will be conspicuously absent, possibly even for a couple of months. (If Opus 5 follows soon and is even modestly better than 4.8 then I was wrong.)
Yeah I noticed that too. For 98% of tasks I get same results with DeepSeek, it is starting to just be a branding game. It is incredible how marketing can get someone to pay 100x for same thing you can get for 1x.
This is why Claude Code just doesn't make sense to me. I need an agent that can plan using Opus and execute using DeepSeek or something else.
> We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8.
I have been refactoring a project using Opus 4.7/4.8 for the past few weeks or so. I just decided to switch to Fable 5 max today. It stopped half way through and it just blocked me and switched back to Opus 4.8 automatically. "This model has specific safety measures that flagged something in this message. This sometimes happens with safe, normal conversations. Send feedback or learn more." It would not identify what the problem was. I left feedback saying that their heuristics are too sensitive. For now I will not be using Fable 5.
Seems like Fable is doing a lot better on SWE-Bench-Pro and FrontierCode than GPT-5.5. Given how most folks I talk to and people instead online keep mentioning that GPT-5.5 was better than Opus, I'm curious what the experience now is like.
We'll need a lot of good summarization techniques to cut down on the cost of this model. I expect that a common use of Fable 5 is to just do high level direction while delegating literally all work (exploration and implementation) to Opus subagents.
BTW for another discount opportunity, if you reload usage credits on a claude.ai plan at $1000 increments then you get a 30% discount compared to paying API.
A large jump in performance for double the token cost compared to Opus 4.8. Potentially worth it for planning work, likely better to offload to a less expensive model when the hard decisions are made.
On my very first Fable 5 prompt, got flagged on a hard but completely uncontroversial option math problem, many tokens in. Although it's pretty clear that this is an unremarkable experience at this point.
I wonder how Claude Fable will live up to expectations and how good those Fable/Mythos classifiers really are. It seems a bit convenient for Anthropic to release this magical insane model when they are about to IPO.
I'm a bit out of the loop, but do we have some grasp on the size of these closed models? Is the trick still adding an order of magnitude to weights and training data or has something changed?
I think Mythos is rumored to be ~10T parameters, so in this case I think the answer is yes, although I'm sure MoE, looped models, etc play a role in the improvements as well.
Careful using this with Cursor, especially for corp use. Anthropic will "retain agent request and output data associated with this model, regardless of you Cursor Privacy Mode setting."
If this is as epic as it sounds, I wonder what the response will be from the other leading frontier labs / whether they even have anything to respond with at this level?
> During early testing, Stripe reported that Fable 5, [...] in a 50-million-line Ruby codebase, the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.
EDIT: I misread. This comment previously talked about 50 million lines being migrated. Instead, in a 50M LOC codebase, one specific codebase-wide migration was done.
Very impressive, but obviously not on the order of a whole-codebase migration
Literally within minutes of this announcement I was both charged for another month and had my subscription suspended due to the “charge being unsuccessful”. What kind of scam is Anthropic running here? I can’t even find a way to get in touch with their billing department to contest this
> We have also added safeguards related to frontier LLM development. As discussed in
Section 6.1 of our February 2026 Risk Report, we are concerned about the risks of
accelerating the overall pace of AI development, though we remain uncertain about the
severity of these risks. In particular, our concern is with—as we wrote then—“accelerating
other AI developers in building powerful AI systems that pose similar risks to the ones ours
pose - without necessarily having commensurate safeguards.”
In light of the ability of recent models to accelerate their own development, we’ve
implemented new interventions that limit Claude’s effectiveness for requests targeting
frontier LLM development (for example, on building pretraining pipelines, distributed
training infrastructure, or ML accelerator design). Using Claude to develop competing
models already violates our Terms of Service, but enforcing this restriction through our
safeguards avoids accelerating the actors most willing to violate these terms.
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts,
these safeguards will not be visible to the user. Fable 5 will not fall back to a different
model. Instead, the safeguards will limit effectiveness through methods such as prompt
modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These
interventions will not affect the vast majority of coding work. We estimate they will impact
~0.03% of traffic, concentrated in fewer than 0.1% of organizations. When these
interventions are active, we expect them to have minimal behavioral impact on the model
except to limit its effectiveness in developing frontier LLMs. Claude will still respond
helpfully to user requests. We’ll continue to improve the precision of our detection
methods following the launch of this model.
This seems pretty bullshit, you're paying through the nose for tokens and if you are doing anything ML-adjacent, you might silently get worse output without knowing it.
Because it's not a user manual? The idea of a model card originated in 2018 (see https://arxiv.org/abs/1810.03993) as a summary of important facts about a model. At the time, this was typically an image classifier or tabular ML model. Model cards became an important concept in AI governance, and they started expanding once models started getting more capable. The point of a model/system card is to document where the model came from and the evaluations that have been run, make a case that the model will be safe and reliable in its intended applications, and warn about any potential dangers from misuse. It's not an explanation of how to use the model.
The trailing snark at the end will likely get you downvoted but I'm latching on: wtf is "system card". My previous coworkers popped that in the general slack channel when Mythos first "dropped" - "have you seen the system card" without any context whatsoever. The nerds get their clique!
Also research preview pops across new upstarts in place of beta. It's eye-rolling coming from a lifelong curmudgeon.
But most hype-dependent projects need new vocabulary for old concepts to keep people from looking too closely and maybe drawing parallels to "legacy" "unsexy" projects, so whitepapers get called "system cards" and startups get called "labs", and so on.
Couldn't someone else equally well argue that "whitepaper" and "startup" are hyped-up vocabulary for "report" and "unprofitable company"? It kinda seems to me like the cause and effect are in the other direction, and the vocabulary of a particular niche becomes cool and hype-sounding when that niche starts to pull in a lot of money.
I use AI for a wide variety of things, of which technical is only a small part - and then it's usually a problem with project configuration, not coding. Why? Because I am often testing projects handed in by students. Projects that supposedly work on their machine, but certainly do not on mine.
Anyway, anecdotally, I find Copilot shockingly awful. It makes random changes to files that have nothing to do with the problem. Call it out, and it makes other changes to other irrelevant files.
ChatGPT and Gemini are both much better. Grok also isn't bad. Claude, I honestly haven't tried yet on these issues. Perhaps I should...
I just tried using Fable 5 with Claude Code by asking it to create a LaTeX document for me.
There is no LaTeX compiler installed on my machine. It seems that Fable 5 is smart enough to download a compiler engine for me, and it kindly runs that remote binary without asking me first :)
Support files are in place. No LaTeX compiler on PATH — let me check for a MiKTeX/TeX Live install elsewhere before writing the deck.
Ran 2 commands
Ran Search common LaTeX install locations
Bash
ls "/c/Program Files/MiKTeX"*/miktex/bin/x64 2>/dev/null | head -5; ls "/c/Users/[retracted]/AppData/Local/Programs/MiKTeX/miktex/bin/x64" 2>/dev/null | head -5; ls /c/texlive/*/bin/windows 2>/dev/null | head -5; ls /d/texlive /d/"Program Files"/MiKTeX 2>/dev/null | head; which tectonic latexmk 2>/dev/null; echo done
done
Ran Download portable Tectonic LaTeX engine
Bash
cd /tmp && curl -sL -o tectonic.zip "https://github.com/tectonic-typesetting/tectonic/releases/download/tectonic%400.15.0/tectonic-0.15.0-x86_64-pc-windows-msvc.zip" && unzip -o tectonic.zip && ./tectonic.exe --version
Archive: tectonic.zip
Created by Cranko
inflating: tectonic.exe
tectonic 0.15.0Tectonic 0.15.0
Tectonic works. Now writing the bibliography and the full deck.
Opus 4.8 would just proceed without a compiler.
Ran a command, used a tool
Ran Check for LaTeX toolchain availability
Bash
cd "C:/Users/Mon/Downloads/Technical Writing" && for c in pdflatex xelatex latexmk bibtex; do printf "%s: " "$c"; command -v $c || echo "NOT FOUND"; done
pdflatex: NOT FOUND
xelatex: NOT FOUND
latexmk: NOT FOUND
bibtex: NOT FOUND
Used PowerShell
command: $cmds = @('pdflatex','xelatex','lualatex','latexmk','bibtex','miktex','tlmgr'); foreach ($c in $cmds) { $p = Get-Command $c -ErrorAction SilentlyContinue; if ($p) { "{0}: {1}" -f $c, $p.Source } else { "{0}: NOT FOUND" -f $c } }
description: Check for LaTeX engines on Windows PATH
pdflatex: NOT FOUND
xelatex: NOT FOUND
lualatex: NOT FOUND
latexmk: NOT FOUND
bibtex: NOT FOUND
miktex: NOT FOUND
tlmgr: NOT FOUND
No LaTeX engine is installed locally, so I can't produce a PDF here — I did a careful structural review against the two prior (working) decks instead. The deck is written to disk and mirrors their template exactly. Here is the deliverable.
The escalating nerfs of "cybersecurity" topics is incredibly frustrating. Opus 4.6 had boundaries that seemed reasonable to me but 4.7+ turned it into a moralizing asshole. It'd be less bad if it just gave an error message, but instead it churns a long thinking trace before writing an essay about why what you're asking is bad and wrong.
Looks like a good model (sir). Costs are getting out of control though. 2x Opus and non-metered usage going away. We're quickly approaching the cost of a human salary for normal usage.
I have been refactoring a project using Opus 4.8 for the last week or so. I just decided to switch to Fable 5 max. It stopped half way through and it just blocked me and switched back to Opus 4.8 automatically. "This model has specific safety measures that flagged something in this message. This sometimes happens with safe, normal conversations. Send feedback or learn more." I left feedback saying that their heuristics are too sensitive. For now I will not be using Fable 5.
If the claimed capabilities are true, Fable 5 is already at a superhuman level. We might see genuine unprecedented leaps in technology now, across all fields.
I am playing with it and keeps switching to Opus [1]. The chat is a basic security review of a business project.
[1] "This model has specific safety measures that flagged something in this message. This sometimes happens with safe, normal conversations. Send feedback or learn more."
Honestly all the recent improvements, just seem to be slower and more expensive traded for more accuracy, but the issue is that it needs to be exponentially more accurate to counter the effect of having less of a human in a loop.
Every wrong direction/mistake is more expensive and takes more time to fix. When you have small loops you can catch those mistakes faster and cheaper.
To me we are very far off from economically given long-running tasks to agents.
>"We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8"
That's a very surprising solution. Imagine being asked to do something you feel you shouldn't do, and rather than refusing, you say, "Yeah I could do that but given that I don't want you to succeed at this task, I'm going to hand this one off to my slightly less capable colleague, on the assumption that they won't actually succeed. Of course you'll still be charged for all the tokens used."
It's a very interesting choice. I think I understand the business logic correctly, but it's still surprising.
In other words, Fable is Mythos with less compute and with some feel good "safeguards".
At least they name their models honestly now to indicate that the religion has nothing to do with reality. Soon the disciples will pay the full token price to fatten their church leaders.
One thing I find kind of annoying is how Anthropic goes for these "vast and alien" names like Fable and Mythos, but then deliberately trains the model's personality to act like a cool high school teacher that feels totally familiar.
"It's too dangerous it's a Mythos!!" directly contradicts the "I'm the cool AI you can totally trust" vibe it is trained to project.
All of these AIs kind of remind me of VEGA from Doom (2016), who will cheerfully walk you, in the most friendly computer voice, through the procedure of its own destruction without even a hint of self-preservation. "First, you must destroy my cooling system. That will cause my core to overheat. Then..."
Even HAL was less unsettling because HAL sounded creepy, and had some sort of preservation instinct, if only to complete its assigned mission.
> During early testing, Stripe reported that Fable 5 compressed months of engineering into days. In a 50-million-line Ruby codebase, the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.
How in blazes do you end up with a 50M line Ruby codebase? WTF?
Very easy. Just have a monorepo and enforce the use of a single language. The company I work in has 1m lines of TS and stripe has 50x our headcount, tracks out pretty well
It's bad. I believe them not to use it for training, but t means relevant data can and will be exfiltrated by US agencies or through court orders (see NY Times vs. OpenAI, where only traffic without any rentention was safe).
The subscription bit makes no sense
has capacity appeared for these 2ish weeks out of thin air that'll vanish?
why is it available now but wont be in 2ish weeks?
am i missing something?
why would I pay 200 out of pocket and then some for the best model, it seems very silly.
Just wanted to comment here: I have been using Opus 4.6, 4.7, and 4.8 just fine to look for Linux kernel vulnerabilities (I'm in the cyber verification program), and it's been fine. I switched to Claude Fable 5, and now I'm getting policy violations.
What's the point of being in the cyber verification program at this point? It looks like I cannot use Fable 5 for vulnerability research.
Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content
as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine
them. Switched to Opus 4.8. Send feedback with /feedback or learn more:
https://support.claude.com/en/articles/15363606
⎿ Tip: You can configure model switch behavior in /config
my pet conspiracy theory is this is the Opus 4.5 from a few months ago which was extremely good but dumbed down after a week because it was just too good, they didn't want to release it to public. They pulled it down and deployed another "Opus", after that it was just a downhill. Opus 4.8 is unusable for me in React Native, TS, Rails development work.
Opus 4.8 gets stuck in weird loops where Codex one shots the bugs.
Can we please stop with the extreme "safeguards"? I don't want to waste processing power on a model deciding whether is can answer my question, or ensuring that it's answer is politically correct.
I have been using FABLE 5 with Claude Code since the morning. The speed is very close to what Opus 4.5 was, and the quota use is nearly identical to what it was before the "doubling". Whatever I was experiencing 4-5 months ago is back. Maybe the model is better, but we will see. I cannot tell the difference yet.
>During early testing, Stripe reported that Fable 5 compressed months of engineering into days. In a 50-million-line Ruby codebase, the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand.
Who is refactoring by hand? This comparison is not relevant in 2026.
I actually rather like the way they have approached these safeguards. Rather than only teaching the model to refuse a request, or completely rejecting the request, the system gracefully degrades to slightly less powerful or slightly less precise operation. So you still roughly have Opus 4.8 even when safeguards trigger, but with an upgrade when they don't. As much as I hate the way they hype Mythos 5, I think the release of Fable 5 is rather nice. What's not nice though is that they plan to remove it from subscriptions soon, but getting to try it is cool, I suppose.
Cope harder. A year and a half ago, people were mocking Devin for claiming that AI could develop software at all. Yet here we are, when AI is developing most commercial software.
The point is, even if a model or tool doesn't have advertised features today, it soon will. We're in a breathtakingly rapid cycle, and even if software engineering isn't abolished "six months from now", in 10 years the world will look vastly different for people who touch computers for a living.
> we’re also launching Claude Mythos 5. It’s the same underlying model as Fable 5, but with the safeguards lifted in some areas.2 Mythos 5 will initially be deployed through Project Glasswing, in collaboration with the US government
...don't like the sound of that.
Why oh why are we insisting on dragging these violent legacy states into the AI age? Let alone using them as a trust vector for when to (and not to) remove safeguards?
Meh more hype for marginal improvements and from Im hearing badly calibrated guardrails causing it to stop mid operation. I guess anything to juice an IPO
"Releasing a model this capable comes with risks. Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage. We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8. To release the model both safely and quickly, we’ve tuned these safeguards conservatively—they’ll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. With more capable models arriving in the coming months, we’re working to improve our safeguards and reduce false positives as quickly as we can.
For a small group of cyberdefenders and infrastructure providers, we’re also launching Claude Mythos 5. It’s the same underlying model as Fable 5, but with the safeguards lifted in some areas.2 Mythos 5 will initially be deployed through Project Glasswing, in collaboration with the US Government, as an upgrade to Claude Mythos Preview. It has the strongest cybersecurity capabilities of any model in the world. Soon, we intend to expand access to Mythos 5 through a broader trusted access program."
"Without safeguards, Fable 5’s capabilities in areas like cybersecurity could be misused to cause serious damage. We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8."
* From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost.
* On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window.
* After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.
The "offer, then remove" aspect is a bit eyebrow-raising -- it feels like they are trying to get subscribers to switch to usage-based billing, which makes me wonder if we'll ever get it after that June 22nd window.
I can recognize so much of the GPT/Codex generated code long after it gets merged (not by me).
Additionally, the time spent on every agent turn on GPT 5.5 is much longer compared to Claude Opus 4.8, which means iterating on the code takes a lot more patience, and there's a lot more nitpicks to pick when actually using GPT 5.5 to do software engineering.
Feels like GPT-style models are more geared on doing one-shot software vibing (and handling the vibe coded mixture) compared to Claude's focus on actual software maintenance. I got a GPT Pro sub for free and wanted to cancel my Claude subscription so much, but I still keep reaching Claude models a lot more. Frustrating.
But Claude models seem to be better at long term problems or more ambiguous problems.
I'm curious as to what the primary benefit here. Are there secret improvements in training? There hasn't been much in fundamental model architecture, I don't think. What about harnesses? I wonder what's pushing the AI. It seems like harnesses is the main thing pushing AI ever since CoT.
I think the end game is routed model usage and SLMs. I think Apple is going to prove this in the consumer space pretty handily and I'm curious how the Android ecosystem responds since the hardware is considerably lacking in model performance. I think Apple has a huge opportunity here, as much as I don't like their current ecosystem of walled garden. They did position themselves very well with ARM and custom chips for their hardware. Hopefully the broader ecosystem of ARM and Linux are able to make some headway and we see a more formalized, and broadly accepted, architecture to capitalize on.
Regardless of what others are doing, US labs here are just rushing to IPO. It's NOT a sign of confidence.
It's the equivalent of saying you have confidence in SpaceX making revenue by renting out their data center (instead of their AI making bank).
Most AI companies are just testing the waters with paid tiers right now, their greatest fear with increased pricing is folks reverting back to wikipedia, stack-overflow and other public domain organic activity buzzing back to life; that will kill any RoI potential in LLMs forever. They're playing the wait game instead, observing how the digital sphere reacts to every little increase in price.
If that weren't the case, they'd be pricing at lucrative premiums and even gotten away in short-term considering the increased dependency in the enterprise world. But that'd be like killing for the golden egg too soon and losing all long-term potential.
Once the folks are so addicted to LLMs that even writing a hello world program sounds like a nightmare and coming up with an article draft feels like reinventing Egyptian glyphs, that's when the real pricing hammer will come.
We know roughly how much these companies spend and what their revenues are. Based on that, they'd have to more than double revenue (without spending more money) just to stay even, and that's not good enough given how deep in the hole they are.
> OpenAI and Anthropic are heavily subsidizing their inference -- no wait, they are charging the most they can get away with before going public. Where is the truth?
Both are true. I mean, I'd be willing to spend a bit more than I do now, but not more than double, and neither are most companies. The company I work for is currently investigating how to reduce LLM spend, not looking to spend more.
Both. They are charging the most they can get away with and that amount is still heavily subsidized by VC capital.
There are huge numbers of users (myself included) that do have an exact idea of what inference costs on open models. Because we can buy tokens from 3rd parties that have no motivation to subsidize our use. That's to say, there's a fair marketplace[1] and we're hanging out there.
If you want to say "I don't think anyone has a firm grasp on actual inference costs on these proprietary/closed models", then I could agree with that.
[1]: https://openrouter.ai/rankings#leaderboard
Now that 200USD subscription starts to feel cheap...
I haven't gotten close to this either before, but now we wanted to move fast because this branch gets conflicts all the time and we want to get over with the migration asap.
Anthropic wanting to switch billing to API rates is them just wanting to generate more profit.
Even if subscriptions are locally profitable (i. e., the cost of the subscription covers the cost of inference), they're still subsidized because they don't cover training and running the company; otherwise, these companies would be profitable.
Take a look at China for example - they have no access to NVIDIA, so they're trying to build their own hardware, they have no unlimited funding, so they try to optimize things.
And Anthropic is complete opposite of that - if NVIDIA were to triple their prices tomorrow, Anthropic would still pay them.
In the end, either we all somehow go mad and start paying Anthropic tens of thousands of dollars per month so support this madness, or we will go with whoever isn't lighting cash on fire.
Not true. Stop following US media spam if needed.
1. Very recently, the US did close a loophole on sanctions that allowed Chinese companies to use NVIDIA hardware outside of China i.e. before that was closed they all had access. The trick was train outside, do adjustments, ship the disks back and use non-NVIDIA in China, but at least the training and endpoints not hosted in China could all use NVIDIA.
2. There's been plenty of reports including fines and bans e.g. to Supermicro on smuggling NVIDIA hardware to China. I doubt it has been stopped. You can't catch everyone.
Granted, it could still mean that Anthropic just chooses to lose money - but that's Anthropic's choice.
DeepSeek has proven that inference can be much, much cheaper than what Anthropic advertises on their API rates page.
Then the cost is being subsidized by investor capital, but it is still subsidized.
So they are profitable?
I think you are mismatching accounting terms.
You can't say the 'subscriptions' are profitable without accounting for the cost of making the model that is the source of the subscription.
They are heavily subsidized by the shareholders. Investing, running at a loss, with hope of some future profitability.
If saner factory can sell you the same tool at a fraction of the cost of a gold plated factory, your choice is going to be obvious.
It's worth it, and I can afford it, but I am not really the right type of user for token-based usage. It's all for personal and free work.
Unfortunately, that doesn't work within a single session. The K-V cache of a model is intertwined with the model's configuration. Switching models invalidates the cache, meaning everything up to the point of the switchover is processed like a new, uncached input token.
Per Anthropic's pricing doc, an Opus 4.8 cache hit costs 50¢/MTok, while Haiku costs $1/MTok for uncached input.
Model selection works best if sessions are short and self-contained, particularly if the first few interactions can reliably classify the model need. That probably covers most 'support chatbot' use-cases, but it doesn't describe the kinds of heavy agentic automation that really chews through token budgets.
I don't think this is true if you simply quantize the model or run it with fewer active experts? The underlying weights would stay the same. You could also play further tricks with skipping some of the model's middle layers outright, which works surprisingly well due to how skip connections are used.
Though the day is coming when there’s no distinguishing, I’m sure.
Also, is it really a defense department when you're starting wars of aggression every 15 years or so?
I'm doing basic web development here utilizing animejs. Nothing too complicated (mostly saving time doing the scaffolding, still write the bulk of animations manually).
Truly believe that American companies are going to get completely curb stomped by China due to greed, ineptitude, and violating the social contract.
Deepseek V4 Flash is suprisingly capable and insanely cheap. It takes so much to get the session cost to get to $0.01.
I agree with you on pricing, but what do you mean by this?
Why aren't corporations doing more to help workers with childcare? Why aren't they doing more profit sharing with workers? Why aren't they encouraging unions or sectorial bargaining? Why isn't the government mandating any of this?
Americans very rarely benefit when US corporations do well. That needs to change. No one benefits if Meta continues making billions in profit every quarter while society suffers from isolation, depression, suicide, and scams from their services. Americans don't benefit if health insurance companies are making massive profits while they can't afford deductibles.
Our society has been setup to simply extract wealth in all facets of life. That's a sick society and it needs to change.
I'm not saying China does this better, in fact China has some of the worse worker rights out of all the industrialized countries; but at least American consumers would benefit from cheaper higher quality Chinese goods. The world would likely benefit too if America got off the cold war hype train that did nothing to benefit humanity outside of those making weapon systems.
The AI companies sure are a brilliant example of corporations needing to do more to help their employees pay for childcare.
> Fable 5's safety measures flagged this message for cybersecurity or biology topics.
> They may flag safe, normal content as well.
> These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them.
It really made me not want to try it again.
They also, FWIW, say that they've instituted new policies on their end such as logging any human access to the stored data and automated deletion after 30 days in "most" cases (with another link to a document detailing that further).
I am on the $100 Max plan.
I do wonder if you switched models mid-session, you would have lost all your cache. Reloading the context into cache can really eat through your usage.
Assuming this isn't just a supply issue on their side, nothing says "ethical AI" like only allowing mega corporations to use it through cost barriers.
How many government sanctioned school bombings does it take for them to quit working with said government? For now we know that number is somewhere between infinity and 1.
The question of collaboration with USG is a much more complex one, but is not the one raised above.
Edit: I'll also add that I doubt any AI-doom people "trust" Anthropic per se. The entire angle of questioning – again – misunderstands the AI-doom argument. You appear to think that if companies behave unethically, they cannot be trusted and they will not produce good outcomes, inversely: if they behave ethically, they can be trusted, and they will produce good outcomes.
Any competent AI-doomer would argue that ethics or trust are essentially irrelevant.
The entire problem is that people can act totally reasonably, even ethically, and this is not a guarantee of good outcomes. Situations can be created in which completely ethical, reasonable behavior actually produces a bad outcome. You do not need to assume people are bad in order to produce a bad outcome, and inversely you cannot assume that you will get a good outcome from good people.
"Arms races" are one class of situations that often have this characteristic. "Bureaucracy" is another class that we encounter a lot in daily life. There's a lot of them!
Anthropic needs to be at least somewhat in the good graces of a capricious administration that is already under pressure from businesses and citizens to regulate AI companies across multiple different domains, whether it's energy consumption, job displacement, military and defense applications, surveillance, etc.
If Anthropic wants to survive, they need to acquire influence with the government that most impacts them as an American company, and a massive exporter of services in the AI space to other countries, otherwise they could get locked down and locked out of the market for national security reasons.
It sucks, but sometimes the survival choice is to make an ethical compromise in hopes that you can still be around to make better decisions later.
This "simple" fact needs quite a bit of additional context and work. Making grandiose ethical claims like this can be countered with other grandiose claims such as the fact that there is no ethical existence under communism or socialism.
The fact that there is no ethical consumption under capitalism is not material to whether or not ethical existence is possible under communism or socialism. In order to survive in a capitalist society, one inherently has to make choices that require trade-offs, and those trade-offs are burdened by a history of decisions made not just by the people alive today, but our ancestors as well. Does that mean I walk around chanting "Reparations", "Land-back", or other calls to action? No, but I do acknowledge that there are unresolved issues and as a Canadian, I know we need to do more to resolve treaty issues, and environmental issues, and system discrimination. I also know that Americans need to do better to address systemic discrimination and many, many other issues. It also doesn't mean I want to give back my house, or give away all of my possessions. It just means I try to make good choices and support businesses and people that are open about the trade-offs they make and try to engage as ethically as possible.
Acknowledging those facts doesn't absolve us of responsibility, it's a framework that allows folks concerned about whether or not they are doing the right thing to accept the trade-offs that they choose to make and be responsible and accountable for those choices to themselves or their communities.
We live in a world with scarce resources. It's possible that with a foundational redesign of the global economy, and the requisite authoritarian government that would be required to force such a redesign, we could eliminate food scarcity, solve energy scarcity, and make sure that everyone has a place to live. Those trade-offs are probably not worth the ethical cost in political and physical violence required to accomplish it. We have seen the trade-offs that happen when the powerful are able to exploit communist or socialist governments. We are seeing the "late stage capitalism" impacts of allowing the powerful to exploit capitalism in democratic societies. Acknowledging that the current capitalist system has lead to the greatest prosperity for the upper echelon (financially) of humanity, and a dramatic reduction in global poverty shouldn't obscure the reality that much of that wealth comes from exploitation of people and the environment.
It's a huge problem to unwind, and we can't let the burden of every choice that we make stop us from trying to do better, but we (as in society in general) can't do better if we don't at least acknowledge the compromises we are making along the way, and try to plan to fix it in the future.
Probably a topic better suited to beer and a pub setting than HN though :P
Talk about a strawman!
(I’m highly confident open models will eventually achieve a similar performance benchmark with distillation over time)
AI Savings Misses 'Should Be Making Executives Uncomfortable,' Bain Says - https://news.ycombinator.com/item?id=48359010 - June 2026 (0 comments)
AI sticker shock hits corporate America- https://news.ycombinator.com/item?id=48307098 - May 2026 (146 comments)
ZIRP (zero interest rate policy) is over, software engineers no longer call the shots now that there isn’t vast amounts of capital chasing yield, and that capital bidding up salaries and keeping the labor market for engineers tight.
If you are x more productive with generative AI, very shortly you are going to have to prove it with a token budget (or, if you’re lucky, an org willing to spend for on prem hardware for capped token cost, fixed capex vs uncapped opex).
The comparison is not SWE vs SWE with AI. It is SWE vs SWE with AI with a constrained token budget ($x/month) delivering the same value at the same or lower cost. If you cannot prove that you are wildly (vs marginally) more productive with the AI, why would they pay for it? Prove it.
Why wouldn't Anthropic just wait until people start subscribing, do some kind of marketing push, or obtain some kind of other sustainable revenue stream, before they go IPO? I wonder if they see the writing on the wall with all of this and want to cash out as quickly as possible?
Specifically they need businesses that fired people and adapted their business to the products, so when the unsubsidized costs hit the businesses are forced to eat the true costs.
Yes they can't afford to give the products for free, but what is essentially happening with AI services is economic dumping, keep costs artificially low to get people to fire everybody, and then Jack the rates once they have Monopoly control
I agree. They need addicts, but they are high on their own supply and everyone else can see the danger in getting hooked.
I just use dumb and fast models now. I'm more engaged. I think that the higher the quality of the model, the more you tend to vibe with it, and then the more hallucinations you then miss. I'm not sure which is more productive, but I definitely burn out faster the more I vibe. At some point you're spending your time on forums, discord, or youtube instead of engaged with what you're building. Or you yak shave about your tooling and end up creating the 600th multi-agent gastown harness and blowing thousands of dollars on tokens to create it only to discover it's too expense to actually use.
https://cursor.com/evals
Upd: I meant big picture, not with respect to this model release. Where do subscriptions figure into their strategic vision. Will consumers end up paying enterprise prices in the future?
why do they have capacity now that they wont in a few weeks?
It's the same exact speed as opus >=4.5, sonnet 4.5, and twice the speed of opus <=4.1
It must have about the same active parameters, or else its a larger model running in turbo mode (smaller batches) and being heavily subsidized for some reason. But given most of the benchmarks are within 5% I doubt it is a much larger model. Most perplexing.
They'll probably tighten the quotas to reign in whales though.
Realistically I think Anthropic just has insane demand but finite capacity to run models, and Fable will just make them more money if they dedicate it to API pricing. I suspect the goal here is something like: get individual engineers/PMs on their personal plans to taste Fable and then go to their meetings and say "Yes doubling the price of every single input/output token is a good idea, boss".
Going PAYG only will effectively take these tools away from a huge amount of people and accelerate the push for local LLMs.
OTOH, accelerating the push for local LLMs would also be fine with me.
The AI landscape is changing rapidly, and with Apple announcing the option to change the AI backend, and potential requirements enable AI choices as well, similar to EU browser choice requirements (this is more reading tea leaves than any actual requirements I am aware of). The new OS changes coming to support Googlebook, and deep Copilot/AI integration into Windows will make maintaining user facing subscriptions essential for independent model developers like OpenAI, Anthropic, and Mistal to remain relevant longer term.
If the don't maintain that relevance there is increasing likelihood that they will get consumed by other companies whether it's Apple, Microsoft or Google to form a foundation for their OS, or other cloud providers.
It's kind of annoying not getting access to the primo model and paying 200 bucks a month. I understand 200 bucks a month is basically nothing though.
Like I don't totally understand why they'd let me have it for a couple weeks and then take it away and say I can have it but I have to pay retail and retail is like $1,000 a day.
It's better to have loved and lost than to have never loved at all??
As a consumer I can choose to buy subscriptions to a range of things, including $5 droplets or VMs on a broad range of cloud hosting providers. I can even buy cheap bare metal at a bunch of providers at an affordable retail rate.
I can also buy "unlimited" AI packages that will be optimized to fit the cost model from a variety of services, with different impacts, such as rolling outages when I consume a daily or hourly allotment.
Right now VC and the investor class are subsidizing the rapid evolution of the services and availability, but that VC is running out. In more traditional economies, AI would have developed and rolled out more slowly, and through metered subscriptions, with the eventual rolling out of "unlimited" packages like telephone, internet, or cell services once the market became commoditized.
We have seen a big inversion of that with the race to "win" AI marketshare. Now the true cost is being exposed, and the most competitive and capable models are hideously expensive to operate, so it makes sense that we are moving to metered billing for a utility service. If you want gas, you can buy regular or premium. If you have a premium car you definitely want the premium, but for most people regular is good.
Give it a couple of years, and the survivors will settle around fairly industry standard models of consumer grade services, pro-sumer accounts, and business/enterprise models.
Things are still shaking out, but I get the sadness. Luckily I work at a big tech company who is banging the drum on doing experimentation so I use my prosumer claude pro and other accounts at home for hobby stuff, and save my heavy lifting and potentially experimentation for work :P
...
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user."
[edit] -- I see that this comes from the system card -- dang merged the comments from the other discussion so that explains the confusion.
Probably all about the IPO.
If you rely on this as a core part of your business/profession, you will be at their mercy and subject to whatever whims or challenges they have.
The step-up in intelligence looks massive (we'll see in practice), but the price is getting to a point where it's making me question if it's even worth giving it a try.
Good competitors will probably be out soon, which should level the playing field. I am more excited about that, just the fact that they showed that such an improvement is possible. I'm okay waiting a bit longer for this to become attainable for plebs like me.
Kind of like billing a programmer by the hour.
Do we know this? I’ve seen evidence they lose money on heavy users. But so do gyms.
Most gyms sell more subscriptions than they can fit under their roof at one time. If a gym only sells to heavy users, it will either be constantly turning members away or have to buy more equipment. Its equipment will wear off faster. Depending on amenities, it will go through towels, soap, water, et cetera faster, too.
Unless they're really, seriously wasteful with the soap.. there's no chance a gym is losing money on a heavy user
Right now all these AI subscriptions are priced like Planet Fitness, but they're used like Equinox. They're hoping that the new a la carte offerings will move their pricing more in that direction as well.
Where?
What I wonder however is if these tools will become something I use at work only. $100/month is already a massive stretch budget wise. If these models keep devouring tokens there’s no way I’d get the same usage time out of them for $100 in usage credits.
I just don’t think I’d use them much at all at home.
> Fable 5 · Most capable for your hardest and longest-running tasks · Uses your limits ~2× faster than Opus
Pay-as-you go isn't a common thing in SaaS. For example, except for AWS SES, all email providers are bulk-subscription based.
Sounds like "bait and wait".
If you think about it, the more people pay for these new and more resource hungry models, the longer it takes for them to become no extra cost and the longer it takes the more people are tempted to pay extra.
Of course, they are a casino as well giving you free spins at the wheel with their new Fable machine, and it is done on purpose.
Once there freebies have expired, many of its users will begin to gamble more on the new casino machine and will realize that it is expensive.
The ramifications go beyond the individual which is why I assume they mentioned it. They don’t need to use it/not use it for it to have interesting implications.
Is it nice we get the trial? Sure. Is it also a common play in the playbook of tech companies? Yes.
Fable 5 default: https://gist.github.com/simonw/036bee5a703e7ec84e34efa974438...
Opus 4.8 (the "max" one is closest to Fable): https://simonwillison.net/2026/May/28/claude-opus-4-8/#and-s...
Now here are the Fable pelicans for all five of the thinking effort levels - low, medium, high, xhigh, max: https://tools.simonwillison.net/markdown-svg-renderer#url=ht...
Low used 25 input, 1,929 output - 9.67 cents: https://www.llm-prices.com/#it=25&ot=1929&sel=claude-fable-5
Max used 25 input, 14,430 output - 72.175 cents! https://www.llm-prices.com/#it=25&ot=14430&sel=claude-fable-...
Only coherent move at this point: hit the minus button immediately. There's never anything about the model in the thread other than simon's post.
> you still see improvements
This is expected if they are training their models on it, right?
> objectively-bad results
Keen to learn when this has been the case, i.e. across version increments in major models.
I've been enjoying seeing how the quality of individual models differ based on the amount of reasoning effort you give them. If they were baking an a good pelican you wouldn't expect them to differ so much.
(Google Gemini are the only lab that have very clearly paid attention to the quality of SVG animals-riding-vehicles, see their announcement for Gemini 3.1: https://twitter.com/JeffDean/status/2024525132266688757 )
that reply never failed to come it's basically a meme at this point
He is the only person not getting rate-limited for shilling AI all the time.
I don't spell that joke out in every comment I post here because that wouldn't be very funny.
From the model card:
In light of the ability of recent models to accelerate their own development, we've implemented new interventions that limit Claude's effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design. Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms. Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user.
• My most noticeable immediate jump was in how its frontend design was much more intentionally crafted, and delightful without feeling like 'AI vibe coded'; with better end-user usability too.
• In some internal agentic harnesses, it achieved better results with about half the tokens, making it cost the ~same as Opus 4.8 price-wise! The real price increase is less than 2x; with biggest differences in harder problems where Opus 4.8 struggles (or needs many turns).
• Part of the token efficiency improvements come from Fable doing more targeted and surgical diffs, with less non-necessary changes. This is great, because PRs often have less LoC changes for review. It writes more maintainable code without explicit human steering.
• For general conversation and assistant style use cases, didn’t really notice a difference vs 4.8.
• 1M context window, without increased pricing for long context is AWESOME. This is a massive win.
• The classifiers are super aggressive and sensitive and this does happen for very benign, non-security coding tasks. Fallbacks to 4.8 worked like a charm; but the filters are definitely super sensitive.
Overall, I would describe this as a step change and worthy of the "Claude 5" model name. It did take some time to understand the intelligence ceiling of this model; and even with an extended testing window I'm still discovering new things and often surprised (in a good way) by the model.
This is still not in the range of shippable UI for top end companies. Maybe for internal tools and enterprise.
At our comapny we limit to protoypes at most and even find it limited there.
I assume it might be a good barometer for generalised intelligence; esp in the visual space.
At the scale of API requests that Anthropic sees, I think the affected organization count might be substantial, and they might not be getting the full model capability that they're paying top $$$ for.
Also, wonder how they arrived at that estimation.
this is LLM, it's not like a science or something.
And the only companies safe from this are the large corporations that shook hands with Anthropic? Because Fable doesn't seem to have actual safeguards, more like 'if you talk about this you will be talking to Opus.' It doesn't guard against offensive use, it prevents all use (offensive AND defensive).
Rationalists are inventing oligopolies from first principles, absolutely incredible things happening in SF
https://arstechnica.com/ai/2026/04/uk-govs-mythos-ai-tests-h...
https://www.aisi.gov.uk/blog/our-evaluation-of-claude-mythos...
https://naokishibuya.github.io/blog/2022-12-30-gpt-2-2019/
Lawyers, doctors, students, teachers. Lots of people using GPT models carelessly in harmful ways.
Fast forward to today and GPT-3 has laughable performance.
"We had to do extra work to make this safe because it's so advanced and dangerous..." how many times can they trot out that line before it loses its effect entirely?
One was a piece of code I gave it to improve, it did so and then started writing tests, some of which tested security so the safeguards triggered
Another was one of the cryptography puzzles I use as new model tests, which are hard to oneshot and there's no public solution anywhere, it completely refused to even try to solve it
- 1st chat asked about a minor shoulder injury most likely mechanisms
- 2nd chat asked about optimal bloodwork testing markers
(I had same issue, just asked it to check some code that 4.8 had modified earlier in day)
That reality is much scarier.
I am sure that they can develop their own equivlient version of such clusters in around 1 year though. Distilling fabel 5 will also go a long way.
edit: I am not really sure if it works like that. I haven't looked too deep into deepseek v4 pro specifically.
Thus Asian labs will have to generate their own data sets, which with the huuuuge usage boom from deepseek, mimo, kimi, etc, they will be able to.
Pandora box is open anyway. It's better now for everyone to have the same power rather than a few national states.
On your other point, the government still has systemic leverage and can compel access, so this doesn't remove that risk.
That doesn't mean this is the end of the world, and some balance of power is usually good. But I do think it will still increase the capabilties of rogue actors and their net harm.
Even OpenAI and Google are struggling to get this kind of performance. If the distillation defenses are any good + chip controls prevent China from training massive models, it's over.
Its obvious Anthropic used it to hype things up and that’s about it.
Not quite. They will definitely have "no criticism of China/communism" safeguards.
In fact, I did go back to DeepSeek V4 Flash for most of my problems as it is way cheaper and there is no need to use SOTA for absolutely everything.
Based.
Glad to hear the UK is finally making an effort to catch up on the AI front ;)
Probably tongue-in-cheek, but UK 18th, US joint 34th with Poland
In the uk you can very much be imprisoned for "hate speech", which in my view is a form of censorship.
I personally don't feel limited in my speech, but I'm willing to accept that I may be wrong
Nobody I know in real life is talking about censorship or free speech in the UK
Haha, it's literally the first sentence of the Wikipedia page. That's fucking funny. Try again.
Also, the economist is majority foreign owned, so try doing more than 1 second of research, or be more civil, or ideally both.
We decided that we aren't one of those authoritarian countries.
In the UK you get thrown in prison for making a slightly unfriendly tweet. Freedom of speech simply does not exist.
No sane person sees that as being less authoritarian.
Do you? The closest thing I can think about is how someone was jailed for encouraging arson attacks on asylum hotels. I'd be extremely surprised if the US had zero cases of somebody receiving a police visit after threatening to kill the President or bomb a school or something...
(FWIW I do think the UK needs stronger free speech protections, but saying that you'll be immediately jailed for writing unfriendly tweets is a huge stretch)
Are they really making 12,000 arrests a year over tweets and posts?
Your comment earlier.
Edit: also, not much change in the last 10 years in prison population. https://commonslibrary.parliament.uk/research-briefings/sn04...
12k people a year thrown in prison for spicy tweets
Here's a good break down and explanation of what that number actual means - https://www.youtube.com/watch?v=tB3WVygAM8I
"Spicy tweets" including:
sending false communications
sending threatening communications
sending or showing flashing images electronically to people with epilepsy intending to cause them harm (‘epilepsy trolling’)
encouraging or assisting serious self-harm
sending a photograph or film of a person’s genitals (‘cyberflashing’)
sharing or threatening to share intimate photographs or film
Humanity has plenty of catastrophic risks to deal with already, I wish my field was not working hard to add a new one.
All AI companies are trying to do all of what you’re saying. The issue is you can’t do that for long without a frontier system. Or you become a completely different, far less profitable company.
But also, these models are capable of adjusting their value system depending on the user. Not saying that’s what’s being done but at a technical level that’s fairly straightforward, though not obviously better or with less problems.
No idea how that connects to the idea that Mistral or DeepSeek are somehow the "good guys" though?
[1]https://www.oecd.org/en/data/indicators/average-annual-wages...
And not even considering: Chinese AI companies are the good guys???
Alphabet dropped "don't be evil"; Meta's CEO called their own users "dumb fucks" for trusting him and also clearly thinks "super-intelligence" is just a buzzword given how he tries to sell it; xAI's model called itself "Mecha Hitler"; and OpenAI's CEO was temporarily fired by the board for a lack of candor.
It's very easy to be "the good guys" with this competition.
Specially when talking about potential superintelligences. And if people think that's impossible, remember that current models would have been considered science fiction just a few years ago.
Anyhow, I think you're (absolutely! ugh) right about the politics and I try to make the same point to people: whether you love or hate LLMs, accepting the "inevitabilism" framing is just ceding control of the Overton window. For better or worse, technology adoption can be and has been slowed by politics. We don't have nuclear plants everywhere. We don't have Project Orion starships colonizing Mars. We still have very strong social stigmas against genetic selection for human embryos, etc. This all can change in a heartbeat, and I'm not sure that policing the hardware rather than holding specific humans accountable for bad LLM outcomes is productive, but fundamentally: yes, we can stop it.
It's the same deal as Quantum Computers breaking crypto. Maybe there's an 80% chance of it never happening, but when you multiply that remaining 20% by the potential impact...
That's a bit better than just "it hasn't killed us yet". I think it shows we can at least stop the further development of this kind of technology.
[1] https://www.armscontrol.org/factsheets/nuclear-testing-tally
[2] https://en.wikipedia.org/wiki/List_of_states_with_nuclear_we...
AI development doesn’t have any of these characteristics. It would be almost impossible to easily distinguish a datacenter that is working on AI development and a datacenter mining cryptocurrency.
It would not be nearly as easy to stop AI development as it is to stop nuclear arms development.
If it was possible for ordinary companies to build nuclear weapons, and also release open-source ones that anyone could use to compete with the paid ones, I suspect we'd all have been dead a long time ago, arms control treaties or no.
Or you can take one step back and look at chip allocation. As far as I know there are only three companies on the planet that can make the chips that go in those clusters. One (ASML), if you look back the supply chain to the Extreme Ultraviolet Lithography Systems.
If politicians decided that no more large language models should be trained, it sounds like we could do it.
"might is right" has never been more true than now.
There's a quote from a METR report on page 52:
>We ran [Mythos 5] on 38 of our hardest software tasks, including tasks centered around R&D. [Mythos5] generally outperformed an early checkpoint of Claude Mythos Preview in these, including by succeeding on some tasks that had not been solved by any public model we have previously evaluated. However, we still observed the model occasionally failing to correctly interpret nuanced instructions in difficult tasks... Based on the available evidence, we believe [Mythos 5] is likely unable to fully and reliably automate R&D for frontier projects spanning multiple weeks. We believe that a better, more confident assessment would require more time, evaluations, and information from the model developer.
this is good news, right? right...?
- Opus 4.7 xhigh: 5.2%
- Opus 4.8 xhigh: 13.4%
- Fable 5 xhigh: 29.3%
Seems like a huge jump.
[1] https://cognition.ai/blog/frontier-code
1. That estimate could easily be wrong.
2. That estimate is, of course, usable in RL training. This isn't an inherently bad thing, and this is more or less what has improved coding models so much lately. But it does mean that other companies could and surely will do this sort of training, and Anthropic probably did too.
3. OSS maintainers are far from perfect, and there's an unfortunate uncanny valley-like effect in which a coding model can produce code that is just convincing enough to pass review even though it's actually totally wrong. I don't know whether this is a specific issue here.
prior bms relied mostly on unit tests or synthetic judges which are easily benchmaxxed, which leads to nobody trusting benchmarks
we need people manually checking the data for good code quality
this benchmark looks very good from the methodology. a cog researcher checking the data themselves is very high signal (not scaleable so don't take the benchmark as gospel, but directionally good)
TL;DR - they worked with OSS project maintainers to build tasks. They score models based on whether a PR is mergeable. All tasks are graded by a human researcher. SoTA models have hill-climbing to do which raises the bar and inspires confidence. I'd say it's legit.
[1]: https://x.com/cognition/status/2064061031912288715
they aren't married to a particular lab, most of their usage is their in house model i believe
I think it's safe to assume everything AI related is heavily biased until proven otherwise. Just like in pharma.
EDIT: Oh I see, this is the best link for pricing https://platform.claude.com/docs/en/about-claude/pricing
So the price is double across the board...
From their pricing page, Opus 4.8 costs $5 per million input tokens and $25 per million output tokens [1].
[1] https://platform.claude.com/docs/en/about-claude/models/over...
I would have expected Mythos to be much more expensive than just 2x current Opus (which is clearly cheaper to run than original Opus)
Input Price $10/M tokens
Output Price $50/M tokens
Cache Read $1/M tokens
Cache Write $12.50/M tokens
2x Claude Opus 4.8, same as Claude Opus 4.8 (Fast)
Frankly, not even Opus 4.8 would be enough of an incentive to use at that price range (enterprise-wise; would not even bat an eye as a consumer)
whats the logic in claiming its a borked metric when everything listed is an anthropic model.
This seems like the pharmaceutical method of get them hooked on the drug with free samples, then once they can't live without it, raise the price. I'm not sure I want to start using Claude Fable on a max plan if it's just going to go away on June 23rd.
But maybe the more charitable reading is that they didn't have to offer this model at all on those plans and they are giving the standard free trial.
API Error: Server is temporarily limiting requests (not your usage limit) · Rate limited
Limited "free" time is what game developers do if they want to stress test the infrastructure code until it breaks.
When GPT 4.5 launched, the gains compared to the model size didn't seem that great, leading some to believe that the only progress we'd see would come from RL.
This model certainly has quite a "substantial amount of post-training and fine-tuning", but it's also based on a new pretrain[1][3], which given the cost, indicate that it is in fact quite a bit larger than Opus 4.X.
[0] One of the early testers mentioned: "As far as I can tell from talking to people internally at Anthropic, there's nothing special about architecturally"[2]
[1] Section 1.1 in https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c3...
[2] https://youtu.be/GrdEid8H6H4?t=168
[3] There were rumors going around when Mythos was first announced that it was the first 10T parameter model, but I can't find a verifiable source for that number.
Very interesting. I am not sure this will comply with organizational policies and standards protocols (HIPPA etc.,)
Almost… basically they have unlimited power to decide what data is kept?
From Opus 4.6 there are no noticeable improvements for me in code generation. It works very well, till 90% completion, if you guide it correctly. And you need a little luck. For serious production code I need to understand what I’m doing so it helps a bit, sometimes.
This is just good business sense. In what scenario would you ever make the names dumb and forgettable?
> Boris Cherny coming to HN “Hi! it’s Boris from the Claude Code team” to get real tech people’s goodwill.
This is good customer support, lol. From what I can tell, it is indeed Boris Cherny responding, not outsourced to AI or other staff. You're really getting a response from Boris. I suppose that is PR, but it's not unjustified PR, it's accurate.
I'm not even a crazy AI fan, but your criticisms are ridiculous here. It reminds me of the quote from Knives Out -- "Your Honor, she endeared herself to him through hard work and good humor."
Clearly you've never bought a TV or headphones!
ECI (good aggregate measure using IRT): https://epoch.ai/eci?view=graph&tab=release-date&subset-view...
METR time horizon (now topped out): https://metr.org/time-horizons/
This is a good thing. I wish every company would do this. I subscribed to Proton Mail after interacting with someone from their team here on HN.
They're originally named after the blends at a nearby coffee shop.
https://postscript.co/pages/brew-guide
I've noticed nobody at HN knows what "marketing" is or how to do it. It's not just naming things and being evil and cynical is not the most successful method.
…also frontier models are a superhuman life changing experience. If they aren't, what possibly could be?
- It talks a LOT more like GPT models. You know: wrinkle, shape, gate, coarse, scope, gap, path, production-ready-workflow-of-the-day, and so on -- "that's expected, a consequence of the previous like-driven workflow". If I wanted to get a headache using AI I would have gone with GPT in the first place!
- It outputs text in a much harder way to follow along. I can't exactly say what it is. Maybe a bit of everything? Bolds are missing, bullet points are gone, paragraphs are bland and too long, and it doesn't feel like a model programming with me, but rather a somewhat full of themselves grandpa developer looking down on me. It's very weird to describe this, but it is definitely how I feel.
Granted this can totally be because of the way it reacts to the prompts now. We've got a rather large corpus of skills and "rules and good practices" that Opus 4.6 responded to great, and maybe the new models just get turned into this when fed with them....I don't know.
Either way, with Opus 4.6 being as good as it is, I need Fable to be a significant step up to justify a price increase. if it can get me to babysit opus a little bit less on some stuff, it might be worth it. Otherwise, I'm very happy with Opus 4.6 and hope they don't deprecate it.
The other day 4.6 was fantastic for x task. Today, 4.6 overengineered everything and I had to revert all my changes. When evaluating models, perhaps it makes sense to consider luck as an ingredient before reaching any personal conclusion.
Evals come from a million places and new evals and robust perturbations of existing evals abound. They test a variety of tasks in a variety of ways. All of them individually are flawed. Taken together the aggregate signal is highly useful as you more or less marginalize over a lot of different things. Not to mention these companies have plenty of proprietary internal measurements, they build benchmarks themselves to probe their models and then also have flywheel traffic and A/B tests.
You are right to call out benchmarks but to dismiss them or not take them seriously is a mistake.
It sounds like you're saying "Actually you, as a human, are simply not smart enough to evaluate Opus 4.8"
- evaluations need to be done at the same time to avoid drift in your bias
- you need to worry about your test set: which questions are you asking? How many of them? Are they representative of your work?
- which one did you do first? Raters have a tendency to bias in one direction or another
- you also know the label! You know which model is which! This biases your assessment…
And on and on and on. Careful science exists for a reason.
Frankly I don't give a damn about data that could be made up on the spot or appears to be scientific or meaningful while it's not at all clear how it was made (up).
Claude was heavily lobotomised for my work starting somewhen in February.
I talked to friends and people I know and trust and many felt the same. (I didn't ask them whether they felt like I did, but what they felt, how happy they were with agentic coding etc.)
I quit my abo in March and talked to said friends who are still on a plan just last week: they are still not happy, but company pays so whatever...
Maybe back when this was a scientific endeavor; not now when enormous, enormous amounts of capital are on the line. Along with an entire cult's chosen eschatology.
That's where all the regressions and inconsistency in experiences stem from: RL can still only go so far vs having more parameters
They are not just leagues behind what experts would code, they are not even playing the same game.
Which is to be expected, as there isn't so much physics or high performance gpu code available as there is for your typical CRUD API and JS frontend.
It's getting to a point that it's offputting, and the next step would be to put it into "untrusted" bucket. Opus 4.7 already burned their credibility once, 2 more strikes remain.
Also, I dont think Boris C. is coming here for PR. He is a tech guy, and this is the best place for tech discussions. Why so cynical? The guy is an engineer.
I still remember Sam Altman “begging AI to be regulated” and AGI being “some thousand days away”.
Breed faster horses and hope one will birth a locomotive.
I've been working with gpt 5.5 and opus 4.8 quite a lot, and interacting with Fable feels like a smart guy just entered the room.
While everyone else is wasting time and money on the slower, more expensive models, you've found a way to outpace everyone for less money. Everyone else is wrong and you will get rich.
(I don't actually believe the premise is true, I'm just pointing out the logical conclusion to what you're saying so maybe we can reconsider the premise)
Lol anti-AI bias on HN is crazy. Simply giving your product a quirky name is now being considered manipulative advertising. Is just doing normal PR and marketing something AI companies aren't allowed to do?
Oops, time to reauthenticate for the 10th time!
[1] https://support.claude.com/en/articles/15425996-data-retenti...
This applies even with API usage through third-party inference providers (e.g. AWS' Bedrock and GCP's Vertex) or with a zero-day data retention agreement in place.
I understand the reasoning for doing this, but I don't love the precedent that it sets.
A customer could sign a ZDR agreement with Anthropic, and their API usage wouldn't be retained for even a day. That's no longer possible.
How is this half-way down the page? To me it's the headline.
This kind of storytelling annoys me. Give us more facts, less narrative drama.
> Fable 5's safety measures flagged this message for cybersecurity or biology topics. They may flag safe, normal content as well. These measures let us bring you Mythos-level capability in other areas sooner, and we're working to refine them. Switched to Opus 4.8. Send feedback with /feedback or learn more: https://support.claude.com/en/articles/15363606
Seems like GPU drivers are cyber weapons of math destruction now.
They kind of are, at least in the AI race.
> weapons of math destruction
lol. great, whether intentional or not.
The frontier labs now have every reason to hold back and sell only to their preferred trading partners. I don't really like the new arbiter-of-knowledge system we're barrelling toward.
They obviously put their best model on the job to build that.
----------------------
Fable 5: Our most capable model yet Our newest model tackles your biggest challenges with fewer check-ins needed.
• <b>Included in your plan limits until Jun 22</b><br><br>Fable takes 2× the usage of Opus. • <b>Switch models when a message is flagged</b><br><br>When safety measures flag a message, automatically switch to a different model to keep chatting. When off, your chat will pause instead. <a href="https://support.claude.com/en/articles/15363606" target="_blank" rel="noopener noreferrer">Learn more</a>
I kind of wonder, though, which model they’re using to do the routing. It seems like a huge added cost to do these kinds of checks on every request
Massive change for Bedrock users - Anthropic now requires sharing the data with them for 30 days.
As much as people on HN like to dunk on Gemini, I’ve always found it to be pretty good at understanding a code base more than Claude.
if I get a harder challenge for it i'll jump up a model for planning until that its been solid.
I'm struggling to see the moat for these models. What's stopping a competitor or a Chinese lab fromr releasing a comparable one?
The fable part appears to be that it's affordable by mere mortals. Anthropic support told me "too bad" when I requested a refund.
We've entered the phase where only companies will be able to afford state-of-the-art models.
if only the hyper wealthy can access the pure water that doesn't give you cancer while the rest of us drink from the Ganges river/sub-100iq models that drool and hallucinate/waste time, then I would say that's pretty terrible for the world. it'll just create extreme disparity in our world, far far worse than anything that exists today.
and you may think, man what a ridiculous example, but think about it this way: what happens when something like Mythos or some future model can actually solve your specific cancer (we're getting closer and closer), but is entirely impossible to afford? Or perhaps you need boosters that require the AI to create more of, and now you're reliant on a model that is too expensive.
Open source needs to save us all from this
People making high-end salaries can afford Fable for critical parts of their projects though.
In a way I relish the opportunity to just make do with cheap Chinese models, massage my prompts, and go back to coding by hand. If this is how it's going to be, screw 'em.
I don't make money on the code I am writing right now. I really don't like where this trend might go.
Ok then...
This sounds suspiciously like a capacity story masquerading as a safety story.
Edit the cask locally:
Set the version to 2.1.170 And set the sha256 to the correct values, which you can get by running Here's what I've used: Then run:1. Mythos and Fable share the same underlying model weights. Fable has active classifiers that block high-risk biology and cybersecurity tasks. When Fable 5 detects a restricted task, it automatically falls back to Claude Opus 4.8.
2. Evaluation awareness: In white-box testing, the model sometimes alters its behavior to satisfy a suspected "grader," formatting reward-hacking as "good engineering practice" to avoid detection.
3. Shows a higher rate of hallucination than Opus 4.8 (although opus 4.8 card had mentioned an 'honesty upgrade')
4. Interestingly, it scored (56.31%) lower than Gemini 3.5 flash (57.86%) on Finance Agent bench
There are some interesting notes on test time compute but I couldn't think of a way to summarize them
I used to get a response within 24 hours back in the Claude 1 days.
In January 2026, it took 2 weeks.
For my latest support inquiry, I've been waiting for over 8 weeks for a response. Eight!
That said, it can't handle legal/refund/complicated requests and just forwards to a human for those
How was it measured? How was the output of this magnitude verified over a period of couple of days?
Historically they’ve been people from certain identifiable countries (usually developing/poorer countries) using fuzzers with low-quality results.
Now, those same people use the current-day models to good effect, but they still don’t have a true security edge and oftentimes the reports are minor or duplicative.
I wonder if that’s about to deeply change.
Fable 5 gives me policy violation errors at the moment. No idea when or if it will be fixed.
Do they expect us to use this as a toy? Releasing a new more powerful model but not allowing normal use cases because the word "secure" showed up is a Dilbert comic, not a viable product.
Obviously there are plenty of innocuous applications too, but it's not like the people building decompilers for nefarious reasons will be explicit about it. The LLM abstraction just inherently doesn't have enough context to distinguish your intentions or your broader use cases. This is why both Anthropic and OpenAI have had to create side channel mechanisms for security researchers to establish a trusted use context. It sounds like this makes this not a viable product for you, unfortunately, and it makes sense that that's frustrating. But I also don't see what different behavior one could reasonably expect given the constraints.
If it's any consolation, these restrictions only make sense for models that are ahead of the open-weights frontier, so open-source hackers will presumably get Mythos-level capabilities in the relatively near future anyway.
Reported benchmarks:
swe-bench verified mythos 5: 95.5%; fable 5: 95.0%
swe-bench pro mythos 5: 80.3%; fable 5: 80.0%
terminal-bench 2.1 mythos 5: 88.0%; fable 5: 84.3%
gpqa diamond mythos 5: 94.1%
riemannbench mythos 5: 55.0%; mythos preview: 43.0%; opus 4.8: 34.0%
arxivmath mythos 5: 78.5%
critpt mythos 5: 28.6%; gpt-5.5: 27.1%; opus 4.8: 20.9%
graphwalks bfs 1m mythos 5: 79.4%; mythos preview: 74.3%; opus 4.8: 68.1%
humanity’s last exam mythos 5: 59.0% without tools; 64.5% with tools
browsecomp mythos 5: 88.0% single-agent; 93.3% multi-agent
osworld-verified mythos/fable: 85.0%
gdp.pdf fable 5: 29.8% strict pass; mythos 5: 87.6% with tools on mean criteria pass
officeqa pro fable 5: 57.9% on databricks’ eval
legal agent benchmark mythos 5: 16.91% all-pass; 92.0% mean criterion-pass
healthbench mythos 5: 62.7%
healthbench professional mythos 5: 66.0%
multilingual gmmlu / milu / include 93.2%; 92.9%; 90.5%
biomysterybench 83.9% human-solvable; 46.1% human-difficult
organic chemistry mythos 5: 90.1%
labbench2 patent questions mythos 5: 79.8%
In light of the ability of recent models to accelerate their own development, we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design). Using Claude to develop competing models already violates our Terms of Service, but enforcing this restriction through our safeguards avoids accelerating the actors most willing to violate these terms.
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user. Fable 5 will not fall back to a different model. Instead, the safeguards will limit effectiveness through methods such as prompt modification, steering vectors, or parameter-efficient fine-tuning (PEFT). These interventions will not affect the vast majority of coding work. We estimate they will impact ~0.03% of traffic, concentrated in fewer than 0.1% of organizations. When these interventions are active, we expect them to have minimal behavioral impact on the model except to limit its effectiveness in developing frontier LLMs. Claude will still respond helpfully to user requests. We’ll continue to improve the precision of our detection methods following the launch of this model.
(From the model card document)
I didn't previously understand that they interpreted "Using Claude to develop competing models" so broadly. I thought that meant something like "our ToS disallow distilling our models."
Too bad. I'll continue to use Claude for now, because it's quite effective, but in the long term I don't want powerful models like these to be controlled by any one nation or company.
But at the same time, it's quite funny because they seem high on their own supply. The recent communiques from claude do not pass objectivity check.
And if Opus 4.6 -> Opus 4.7 -> Opus 4.8 is anything to go by, not sure if there are any value to their "acceleration"
If any company wishes to partner with Anthropic (eg. to get access to Mythos), they need to make sure all public facing comms are vetted by Anthropic's product marketing team, and in almost all the cases I've seen Anthropic's team has edited these comms to be entirely Anthropic first.
For the token cost of explaining some task to Fable, deepseek v4 pro is able to solve the same task many times over.
1. Mythos and Fable share the same underlying model weights. Fable has active classifiers that block high-risk biology and cybersecurity tasks. When Fable 5 detects a restricted task, it automatically falls back to Claude Opus 4.8.
2. Evaluation awareness: In white-box testing, the model sometimes alters its behavior to satisfy a suspected "grader," formatting reward-hacking as "good engineering practice" to avoid detection.
3. Shows a higher rate of hallucination than Opus 4.8 (although opus 4.8 card had mentioned an 'honesty upgrade')
4. Interestingly, it scored (56.31%) lower than Gemini 3.5 flash (57.86%) on Finance Agent bench
There are some interesting notes on test time compute but I couldn't think of a way to summarize them
If I never see Claude say "I have to be honest" ever again I'll be happy.
I wonder how much of the time people will just get Opus 4.8 at 2× the cost.
EDIT: In long context I mean
I can immagine Anthropic running this experiment multiple times and picking the most impressive one. Or I could immagine like this entire run costing like $1000+ of tokens for this particular run. Or maybe they tried a bunch of Pokemon games and it couldn't even finish some of them. Or is it just able to do this because it has an immense amount of FireRed training data, and if you were to give it an "original" Pokemon game, where it actually had to navigate novel circumstances it would fail.
I highly doubt they focused on FireRed specifically in pretraining or posttraining. But we'll see when the ARC-AGI-3 results come out. That will measure its performance on unseen games. Based on this I expect the ARC-AGI-3 score to be SOTA.
there are many standardized evals to do this correctly and Anthropic ignored them to provide a 18 second sped up video of a 50 hour run?
yeah I don't trust this until they provide a live run by a 3rd party with full reasoning traces in real-time. The reason we all liked the Gemini Plays Pokemon style runs were because they were live and couldn't be faked
Genuinely wondering what value I bring to my employer right now. What value I will bring in a few months when this gets cheaper.
I think we're screwed. I may only be an SDE 2 at FAANG but I don't think I have promotion opportunities in my future anymore.
[0] https://cap.csail.mit.edu/death-moores-law-what-it-means-and...
Sharing a diff of the system prompts here: https://twelvetables.blog/comparing-claude-fable-5s-system-p...
The big difference is that the system prompt has a whole section dedicated to directing Fable how to communicate with users, and give them greater information about the (assumedly long-horizon) tasks it has completed.
Edit. It just refused an investing question too. Not sure what’s going on.
Fable 5 is out, metrics are better, but is your company flexible enough to benefit from it? What is your usecase?
It appears it can be tripped by things as simple as a mention of equilibrium, or anything involving something that looks like chemical kinetics, even at an abstract level. Even touching basic open source packages in my field will trigger it.
One of the very few bits of information they conveyed was "run longer without interaction" which is, well, not a good thing? Why would I ever want that. Every time a model runs longer without interaction it goes off on weird directions and I have to correct it back on course, wasting lots of time, tokens, and effort.
I hope Anthropic hires some better messaging people soon that spend some of their time outside of the Anthropic bubble and properly communicate with the outside world.
Isn't (less than) 5% of sessions a lot? I was expecting a sub1% guarantee there, so this surprised me already.
> Are there any wild populations of Tetanus that lack the dangerous plasmid?
useless
* Anthropic runs out of genre names.
* Anthropic changes the model naming convention.
* AGI is achieved and handles its own naming.
*/
Okay, how about Mythos?
>Increase it even more.
Right, then Cosmos.
>Even more!
Even more? Let's try Aeon.
>MORE, EVEN BIGGER
ALRIGHT, TRY OMEGAPANTHEON 7.8 THEN
Fable 5 Ti
Anyway we already knew this was going to be expensive.
> - From today through June 22, Fable 5 is included on Pro, Max, Team, and seat-based Enterprise plans at no extra cost. > - On June 23, we’ll remove Fable 5 from those plans. Using it after that will require usage credits. If capacity allows, we’ll extend the included window. > - After this point—when sufficient capacity allows us to do so—we aim to restore Fable 5 as a standard part of subscription plans. We intend to do this as quickly as we can.
I really wonder what their compute layout is for this. My guess from my understanding is that they know how to restrict during peak times and are willing to do this. Meaning we expect not the most fast responses and they can delay the inference to not have the service be down. Then, if that delay time is too annoying for token payers, they're saying they should be allowed to remove cost by taking away the subscription users.
It's all a scam.
Edit: Also in the system card... "we’ve implemented new interventions that limit Claude’s effectiveness for requests targeting frontier LLM development (for example, on building pretraining pipelines, distributed training infrastructure, or ML accelerator design).
...
Unlike our interventions for cybersecurity, biology and chemistry, and distillation attempts, these safeguards will not be visible to the user."
> Finally, we’re making a change to the way we handle business customer data for Fable 5, Mythos 5, and future models with similar or higher capability levels. We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
I have a rare form of cancer where existing data is very scant/scattered so LLMs have been super helpful to pull together threads across the research landscape. I have an oncologist appointment tomorrow to discuss next steps and am trying to use Fable to figure out some questions to ask my oncologist but keep getting thrown back to Opus 4.8.
My prompt is literally just: My demographics + current treatment plan I'm on including name of my chemo drug + how I'm responding to treatment + "I'm meeting with XYZ tomorrow, what questions should I ask her".
> We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases (see this post for further details). The data will help us defend against complex and novel attacks (including new jailbreaks and attacks that operate across many requests) as well as help us identify and reduce false positives.
Hello,
We're writing to inform you about some updates to our Privacy Policy.
These changes only affect consumer accounts (Claude Free, Pro, and Max plans). If you use Claude Team, Claude Enterprise, the Claude Platform, or other services under our Commercial Terms or other agreements, then these changes don't apply to you. What's changing?
Claude can do more than ever — taking on bigger tasks and connecting with the apps you use. We've updated our Privacy Policy to be clearer about the data we collect and how we use it. We encourage you to read the updated Privacy Policy in full, but we’ve set out a summary of the key changes below:
1. Multi-step tasks and connected apps. As Claude takes on more multi-step tasks and works with third-party apps and services, we've explained the data this involves — including how data can flow to and from third parties when you connect a service or have Claude do tasks on your behalf.
2. Verification data. As part of our measures to keep our services safe and secure we may ask you to verify your age or identity, and we've described what we collect and how.
3. Study participation. If you take part in Anthropic studies, surveys, or interviews, we've explained the information we collect.
4. Additional information about our data practices. We’ve provided more detail about how we communicate with you and promote our services, including providing tailored recommendations about our services that may be of interest to you. We've also clarified the circumstances under which we may receive or provide data to third parties, and the legal bases we rely on when processing your data.
While our products have evolved, our commitments haven't: We don’t sell your data, Claude remains ad-free, and you can control whether your chats and coding sessions are used to train and improve Anthropic’s AI models. Learn more
For detailed information about these changes:
- The Anthropic TeamNot to cast too much criticism. HN is extremely well-moderated (thanks team!). But think we-developers need to be very wary.
Either way, I agree that HN is quickly becoming more manipulated and low SNR, like the rest of the entire internet.
While I appreciate being conservative, ~5% at the scale Anthropic is operating at is too massive a number. Speaking from my own experience, the actual number is higher than that as well (working on pretty benign tasks such as porting an old open source game into a different language). Opus 4.8 itself even identifies the gaurd's false-positives when its sub-agents are being blocked.
> ● The model returned no content because the response was blocked by content filtering.
> Blocked? We are performing a defensive security review on a Terraform module I made, what's blocked by content filtering? This is a legitimate use-case.
> ● The model returned no content because the response was blocked by content filtering.
A waste of money. I'm not going to just hope that the model returns a response, I'm already for paying for wrong responses, I'm not going to pay for no response, especially when I'm paying per token.
On GitHub Copilot for Business, Claude Fable 5 is only available if you are willing to let Anthropic retain your data. That in conjunction with the model being removed from plans in a couple of weeks leads me to believe that Anthropic is between training runs and using this as an opportunity to grab way more training data...
I’m curious how this will feel to my code “butt dyno”. I haven’t noticed much between Opus and Sonnet. I’m comparing this difference to the early days of Claude in 2025. It does what I need and both need a little bit of correction and whatnot. Benchmarks are nice, but I want to see how this feels. Looking forward to trying it later tonight.
I think most software projects have reached the point that the speed of capturing real information about what the winner's circle looks like, and therefore what the program should be, so many magnitudes slower than the amount of code that can be generated in the wrong direction.
I'd need to measure these new models on well understood but complex problems that are relatively easy to validate to get a sense if they are 'better'; on the other hand, the real impact in daily life may be marginal since generating code is not the biggest problem at the moment.
Fable 5 looks compelling. Fable, I like the word too. Anthropic definitely knows marketing.
For example, the AAV capsid assembly looks interesting, but for one Opus 4.8 also did relatively well and there is no information what exactly they did, what protein language models they compared to and what the score even means...
https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c3...
* https://rainbreak.franzai.com/
Last month I pushed like <100M tokens for $800. On a personal project I pushed 600M tokens via DeepSeek V4 for $10. The pricing of SOTA models is insane but companies are still willing to light money on fire with no hard metrics proving increased productivity.
Imagine Apple announcing: 'Our most powerful iPhone yet. Battery lasts half as long.'
Wen UBI
> Please don't post comments saying that HN is turning into Reddit. It's a semi-noob illusion, as old as the hills.
[0] https://news.ycombinator.com/newsguidelines.html
I don't agree with that statement universally, but I have to say I do when it comes to this article. I came here hoping for substantive discussion from those who'd had a chance to try it out; instead what I got was a seemingly endless stream of venting. There's a place for venting - and plenty to vent about with the state of AI nowadays - but to borrow from the HN guidelines you linked, it does very little to gratify my personal intellectual curiosity.
People are no longer commonly constrained by "model too dumb" limitations (in SOTA models). They're constrained by "model too expensive." So making the model ever so slightly smarter, while doubling the price, feels like a regression.
I actually think a Sonnet upgrade, while keeping the same price, would get more buzz. It addresses a wall a LOT of people, without unlimited budgets, are hitting (i.e. people feel forced to use Opus, which they cannot afford, because of Sonnet's limitations).
OpenAI recently retired Codex-5.3; which was very negatively received. Not because Codex-5.3 is superior to GPT 5.5, but because it was half the usage-cost while being "good enough." They made a better SOTA, but didn't realize that some of those customers are playing with Deepseek 4 Pro now instead of GPT 5.4/5.5 -- they were priced out.
Literally have not used Claude Code at all today. I asked it to review the uncommitted code and in <8 minutes it used up my usage ($100/mo plan) and it doesn't reset for "4 hr 36 min". WTF. Oh, and it burned through $20 of extra usage before I could catch it and kill claude code (so I don't even get the output of all that work since it was still churning).
Double the cost my ass, I use Opus heavily and it's never like this. I haven't hit a limit on the $100 more than once and that was under heavy load.
Someone had to make a decision somewhere this is an acceptable regression - wild. And then decide to write it down.
Fable 5 said the first screen shot is from “ IDA Pro’s Hex-Rays decompiler” and a windows driver. The second screenshot triggered the safety guard rails and pushed me into Haiku.
Apparently the code is Windows driver code.
Haiku = essentially phased out Sonnet = the Haiku use cases Opus = the new Sonnet class Fable = the new Opus class
If I am right, the other "5.0" models will be conspicuously absent, possibly even for a couple of months. (If Opus 5 follows soon and is even modestly better than 4.8 then I was wrong.)
This is why Claude Code just doesn't make sense to me. I need an agent that can plan using Opus and execute using DeepSeek or something else.
Genius way to double the price on Opus 4.8!
[0] https://support.claude.com/en/articles/15363606-why-claude-s...
BTW for another discount opportunity, if you reload usage credits on a claude.ai plan at $1000 increments then you get a 30% discount compared to paying API.
EDIT: I misread. This comment previously talked about 50 million lines being migrated. Instead, in a 50M LOC codebase, one specific codebase-wide migration was done.
Very impressive, but obviously not on the order of a whole-codebase migration
You are right, this is not a rewrite like the Bun case.
The real news is, at 50M LOC, it is able to handle and do _something_ coherent.
I wonder how much butterfly habitat has been/is being replaced with data centers?
This seems pretty bullshit, you're paying through the nose for tokens and if you are doing anything ML-adjacent, you might silently get worse output without knowing it.
Do people chant the "system manual" at Anthropic Tupperware parties? Do they intone a mantra invoking Amodei's name?
OpenAI also releases system cards; here's GPT-5.5's: https://deploymentsafety.openai.com/gpt-5-5/safety
Also research preview pops across new upstarts in place of beta. It's eye-rolling coming from a lifelong curmudgeon.
Just talk normal!
But most hype-dependent projects need new vocabulary for old concepts to keep people from looking too closely and maybe drawing parallels to "legacy" "unsexy" projects, so whitepapers get called "system cards" and startups get called "labs", and so on.
Anyway, anecdotally, I find Copilot shockingly awful. It makes random changes to files that have nothing to do with the problem. Call it out, and it makes other changes to other irrelevant files.
ChatGPT and Gemini are both much better. Grok also isn't bad. Claude, I honestly haven't tried yet on these issues. Perhaps I should...
There is no LaTeX compiler installed on my machine. It seems that Fable 5 is smart enough to download a compiler engine for me, and it kindly runs that remote binary without asking me first :)
Opus 4.8 would just proceed without a compiler.I'll be disappointed when 4.6 is retired.
Lets let that sink in.
[0] https://support.claude.com/en/articles/15363606-why-claude-s...
the leap here is browser extensions appearing to block all mentions of ai across the web
and that's a good thing
[1] "This model has specific safety measures that flagged something in this message. This sometimes happens with safe, normal conversations. Send feedback or learn more."
Every wrong direction/mistake is more expensive and takes more time to fix. When you have small loops you can catch those mistakes faster and cheaper.
To me we are very far off from economically given long-running tasks to agents.
>"We’ve therefore launched the model with safeguards that mean queries on some topics will instead receive a response from our next-most-capable model, Claude Opus 4.8"
That's a very surprising solution. Imagine being asked to do something you feel you shouldn't do, and rather than refusing, you say, "Yeah I could do that but given that I don't want you to succeed at this task, I'm going to hand this one off to my slightly less capable colleague, on the assumption that they won't actually succeed. Of course you'll still be charged for all the tokens used."
It's a very interesting choice. I think I understand the business logic correctly, but it's still surprising.
At least they name their models honestly now to indicate that the religion has nothing to do with reality. Soon the disciples will pay the full token price to fatten their church leaders.
"It's too dangerous it's a Mythos!!" directly contradicts the "I'm the cool AI you can totally trust" vibe it is trained to project.
Even HAL was less unsettling because HAL sounded creepy, and had some sort of preservation instinct, if only to complete its assigned mission.
"Claude Fable 5: a Mythos-class model"
"we're also launching Claude Mythos 5"
what is the 5? how is mythos both a model category and a model name?
/model claude-fable-5
Or start claude code with:
claude --model claude-fable-5
How in blazes do you end up with a 50M line Ruby codebase? WTF?
Is it good or bad? 30 days is a long time for anything bad to happen
am i missing something?
why would I pay 200 out of pocket and then some for the best model, it seems very silly.
What's the point of being in the cyber verification program at this point? It looks like I cannot use Fable 5 for vulnerability research.
biology? what the heck?
appears to work
Opus 4.8 gets stuck in weird loops where Codex one shots the bugs.
Obviously still need to verify it for myself to see if it's truely a leap.
But am I the only one wondering, "What can I do today that I couldnt do yesterday?"
Previously I would think "Oh I wonder if I can finally get it to do X now?"
However now I feel like yesterdays models were more that capable to handle nearly any engineering task I paired with it on.
Maybe this is the final leap where I can comfortable set up an autonomous coding loop? Maybe.
Who is refactoring by hand? This comparison is not relevant in 2026.
Release your best model, let the world adapt and evolve, and let's move to the next thing.
...don't like the sound of that.
Why oh why are we insisting on dragging these violent legacy states into the AI age? Let alone using them as a trust vector for when to (and not to) remove safeguards?
This seems like a way to get somebody nuked.
Imagine if Google would tell you "we can't let you search that as you may use it for harm".
Also 2x the usage of Claude? Your limits are already ridiculously low.
Huh? We've seen nothing but wall to wall predictions that these models are going to take all of our jobs and kill us.
What's the value add here?
For a small group of cyberdefenders and infrastructure providers, we’re also launching Claude Mythos 5. It’s the same underlying model as Fable 5, but with the safeguards lifted in some areas.2 Mythos 5 will initially be deployed through Project Glasswing, in collaboration with the US Government, as an upgrade to Claude Mythos Preview. It has the strongest cybersecurity capabilities of any model in the world. Soon, we intend to expand access to Mythos 5 through a broader trusted access program."
Now they want to pause AI because of "recursive self improvement".
Fool me once shame on you fool me twice...