Miles Brundage is pushing back on reports that Anthropic's internal Mythos model has delivered a dramatic leap in developer output, framing the system instead as a routine new pretraining effort on par with OpenAI's earlier Spud run and unlikely to confer any unique acceleration edge.
The internal boost from Mythos-assisted development since February is just too big. Anthropic is pulling away from the pack for the first time, and at the same time they are also speeding up. The race legitimately feels like it is changing for the first time in years.
He notes Mythos follows the same pattern as prior base models and expects the upcoming 4.6 release to deliver comparable gains without evidence of outsized internal effects from Mythos itself.
Anthropic's own survey showed wide variation in task-level gains around a 4x geometric mean, yet the company cautions that individual speedups do not translate directly to overall research acceleration once compute and coordination limits are included.
Positive users highlight Mythos performance gains in drug design and talent shifts to Anthropic as validation of leadership, while negative users call the claims shady or unsubstantiated.
Karpathy: 'this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems.'
This is a super exciting release - Claude Fable 5 is the same underlying model as Mythos but with added safeguards. The benchmarks are great and it's SOTA on everything by a margin but I'll add that *qualitatively* also, this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems. You can give it a lot more ambitious tasks than what you're used to, the model "gets it" and it will just go, and it's never felt this tempting to stop looking at the code at all (but don't do this in prod!). The model still has quirks that people will run into and the safeguards are configured to be a little too trigger happy for launch, which can hopefully be tuned over time.
I feel a lot of things changing as working software increasingly comes out on a tap. The Jevon's paradox kicks in and I feel my own demand for software growing substantially. You can ask for anything - explainers, visualizers, dashboards, bespoke single-use apps (e.g. a full wandb that is hyper-specific just for your project), you can 10X your test suite, auto-optimize code, run giant research projects with custom HTML for the results, anything! "Free your mind" (Matrix ref). Really looking forward to all the things people build!
Quotes from the release today:
'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.'
'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.'
'Mythos 5 is our first model to consistently produce novel, compelling scientific hypotheses. In blinded head-to-head comparisons against Opus-class models, our scientists preferred Mythos’s molecular biology hypotheses ~80% of the time, and have advanced several to experimental evaluation. In the meantime, one Mythos hypothesis—a novel mechanism for an E. coli protein—was corroborated in a study from a lab independently working on the same problem.'
'Mythos 5 conducted novel genomics research in over a week of largely autonomous work. It assembled single-cell data for millions of cells spanning 138 animal species and designed and trained a custom machine learning model to identify cells performing the same role in even distantly related organisms. With only high-level human input, Mythos 5’s trained model outperformed a recent model published in the journal Science—despite being 100 times smaller. We intend to publish these results in the coming months.'
Karpathy: 'this is a major-version-bump-deserving step change forward (imo of the same order as Claude 4.5 was in November), peaking especially for long problem-solving sessions on very difficult problems.'
@AndrewCurran_ I don't see any evidence of that. Mythos is a new pretrain. So was Spud. 4.6 will also presumably be good
The internal boost from Mythos-assisted development since February is just too big. Anthropic is pulling away from the pack for the first time, and at the same time they are also speeding up. The race legitimately feels like it is changing for the first time in years.
EM had early access. 'First, how good is Fable? In experiment after experiment I conducted, it outperformed basically every other public model I have used by a considerable margin. It was capable across many problems and produced some startling results — it would work up to a dozen hours executing on multi-page specifications.'
I've had access to Fable for a bit. A genuine jump in capability, I could feed it a 15 page design document for a project and it would work for 9+ hours and deliver terrific results.
But working with it is weird & weirder is coming
Lots of examples: https://open.substack.com/pub/oneusefulthing/p/what-it-feels-like-to-work-with-mythos?r=i5f7&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true
first sparks of RSI
The internal boost from Mythos-assisted development since February is just too big. Anthropic is pulling away from the pack for the first time, and at the same time they are also speeding up. The race legitimately feels like it is changing for the first time in years.
@Miles_Brundage @AndrewCurran_ The problem is that 5.5 was the first Spud release, and it falls flat (even 5.5 Pro) compared to Fable 5.
My guess is that Mythos is a much bigger base model than Spud, so OAI have a lot of catch-up to do with 5.6.
@AndrewCurran_ Not a lot of OpenAI employees tweeting right now in response to Mythos… minor bearish signal for OpenAI
@AndrewCurran_ Curran, you are an excellent follow. The blend of news and takes is exactly what's desired. Enjoy my friend
@Miles_Brundage @AndrewCurran_ I'm assuming you meant 5.6.
My impression is that Anthropic is sandbagging more than OpenAI though. Mythos has existed since February.
I don't think OpenAI has a comparable model they are unwilling to even talk about?
@Miles_Brundage @AndrewCurran_ 4.6? You mean 5.6?
@deredleritt3r @AndrewCurran_ I respect your POV a lot, and I am surprised to read this, so I would love your reasoning behind that when you get it formed
@AndrewCurran_ this feels so sci-fi
@AndrewCurran_ seems pretty shady how they are giving subscriptions only 14 days of access and then will try to reinstate if capacity allows
@AndrewCurran_ pulling a KD move like Secretariat
@AndrewCurran_ They are releasing mythos… unless mythos went back in time and created itself
@AndrewCurran_ Let's fucking go! https://www.youtube.com/watch?v=W8axBpHxZYM
@AndrewCurran_ nah, divine providence belongs to openai, the machine at the end of time is a chatgpt descendent👽
@AndrewCurran_ Do you think Sam will declare code red
@AndrewCurran_ Now they just need to have enough compute
@AndrewCurran_ This is all bullshit. OpenAI could serve the same quality if they wouldn't include it in any plans, but they don't because they aren't immorale greedy fucks that think only the top 0.01% are allowed to have access to advanced intelligence.
Miles Brundage is pushing back on reports that Anthropic's internal Mythos model has delivered a dramatic leap in developer output, framing the system instead as a routine new pretraining effort on par with OpenAI's earlier Spud run and unlikely to confer any unique acceleration edge.
The internal boost from Mythos-assisted development since February is just too big. Anthropic is pulling away from the pack for the first time, and at the same time they are also speeding up. The race legitimately feels like it is changing for the first time in years.