Anthropic Just Released the Most Capable AI Model Ever Made Available to the Public. The Capabilities Are Staggering. So Is the Responsibility.

Claude Fable 5 is Mythos-class intelligence, made safe for general use — and the version they kept back for cybersecurity work is already designing drug candidates and running genomics research with no human in the loop. We are in different territory now.

There's been a slow drumbeat of signals over the past two weeks — recursive self-improvement data, the OpenAI manifesto, the government equity talks, the coordinated pause calls. All of it has been pointing toward a single moment: the kind of model release that changes what the word "AI" means. That moment arrived June 9th.

Anthropic launched Claude Fable 5 — a Mythos-class model made safe for general use — and Claude Mythos 5, the same underlying model with safeguards lifted for a small group of cyberdefenders and infrastructure providers. The capabilities across both are extraordinary. The honesty in the announcement about what that means — and what it costs — is equally notable.

WHAT'S HAPPENING: Months of engineering compressed into days, a game beaten with vision alone

Fable 5's capabilities exceed those of any model Anthropic has ever made generally available — state-of-the-art on nearly all tested benchmarks, 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 previous models. The customer evidence is more telling than any benchmark. 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. On vision: previous Claude models needed complex helper harnesses to play Pokémon FireRed — Fable 5 beat the game with a minimal, vision-only harness, no maps, no navigation aids, just raw screenshots. On memory: in the deck-building game Slay the Spire, access to persistent file-based memory improved Fable 5's performance three times more than it did for Opus 4.8. The model is becoming qualitatively different in how it handles complexity, duration, and context. GoogleGoogle

WHY IT MATTERS: Mythos 5 is running autonomous scientific research — and producing results

The Mythos 5 results are where the announcement crosses into genuinely historic territory. Using Mythos 5, Anthropic's internal protein design experts accelerated aspects of the drug design process by around ten times — with the model, using protein design and bioinformatics tools but no human assistance, matching or beating skilled human operators. It executes all of the tasks 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 yielded strong candidates for drug design currently under investigation. And in genomics: Mythos 5 conducted novel genomics research in over a week of largely autonomous work — assembling single-cell data for millions of cells spanning 138 animal species and designing and training a custom machine learning model to identify cells performing the same role across distantly related organisms. Mythos 5's trained model outperformed a recent model published in Science, despite being 100 times smaller. That last sentence deserves to be read slowly. An AI running autonomous research for a week produced a result that beat a peer-reviewed publication. GoogleGoogle

"Claude Fable 5 works at senior research scientist grade — picking directions, allocating resources, killing its incorrect beliefs, and producing novel first-principles outputs." — Sean Ward, CEO, Imbue

THE BIGGER PICTURE: Safeguards shipped conservatively — and Anthropic said so explicitly

The most honest paragraph in the announcement is also the most important. Releasing a model this capable comes with risks — without safeguards, Fable 5's cybersecurity capabilities could be misused to cause serious damage. The model launches with safeguards that mean queries on some topics will instead receive a response from Opus 4.8. To release the model both safely and quickly, these safeguards have been tuned conservatively — they'll sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. That's Anthropic saying plainly: we know this model is too capable to release without limits, we drew those limits conservatively, and we're going to refine them as quickly as we can. The pricing signal matters too: Fable 5 and Mythos 5 are $10 per million input tokens and $50 per million output tokens — less than half the price of Claude Mythos Preview. Capability is increasing. Cost is decreasing. Access is expanding. All at the same time. Google

MY TAKE: This is what the inflection looks like from the inside

Two weeks ago, Anthropic published data showing Claude writes 80% of its own codebase and called for a coordinated pause mechanism. This week, they released a model that can run a week of autonomous genomics research and beat a peer-reviewed paper. Those two things are not in contradiction — they're the same company doing the most responsible thing they know how to do in a situation where the technology is moving faster than any institution can fully keep up with.

The Stripe example is the most legible signal of what's changed. A 50-million-line codebase migration — the kind of project that consumes entire engineering teams for months — completed in a day. That's not an incremental improvement. That's a step change in what it means to have a software team. The economic implications alone would have been the lead story a year ago. In the context of everything else happening this week, it barely registers as the headline.

The autonomous scientific research results are where I keep returning. Nine drug design candidates under active investigation. A genomics model that beats Science publications at 100x smaller size. These aren't demos. These are production outputs from a system running without human guidance for extended periods. The word "research" used to imply human judgment at every decision point. Mythos 5 is demonstrating that some of those decision points no longer require it.

So here's the question worth sitting with: if a model can now run a week of autonomous scientific research, compress months of engineering into a day, and beat peer-reviewed publications — what does the phrase "human oversight" actually mean in practice, and are we building enough of it fast enough?

Source: Anthropic — Claude Fable 5 and Claude Mythos 5, June 9, 2026