Anthropic Launches Claude Fable 5 and a Restricted Claude Mythos 5
Anthropic released Claude Fable 5 for general use, plus Claude Mythos 5 -- the same underlying model with safeguards lifted for trusted cyberdefenders.
What Anthropic announced
Anthropic says it is launching Claude Fable 5, which it calls a Mythos-class model "that we've made safe for general use." The company says Fable 5's capabilities "exceed those of any model we've ever made generally available," describing it as state-of-the-art on nearly all tested benchmarks of AI capability, with particularly strong results in software engineering, knowledge work, vision, and scientific research. Anthropic adds that the longer and more complex a task is, the larger Fable 5's lead over its other models becomes.
Alongside Fable 5, Anthropic is launching Claude Mythos 5 for what it describes as "a small group of cyberdefenders and infrastructure providers." Anthropic says Mythos 5 is the same underlying model as Fable 5 but with safeguards lifted in some areas, and calls it the model with the strongest cybersecurity capabilities in the world. Mythos 5 will initially be deployed through Project Glasswing, in collaboration with the US government, as an upgrade to Claude Mythos Preview, and Anthropic says it intends to expand access through a broader trusted access program soon.
Why Anthropic built in safeguards
Anthropic is explicit about the risk: "Releasing a model this capable comes with risks," the company writes, warning that without safeguards, Fable 5's cybersecurity capabilities could be misused to cause serious damage. To manage that, queries on some topics are instead routed to Claude Opus 4.8, Anthropic's next-most-capable model. Anthropic says it tuned these safeguards conservatively to ship both safely and quickly, acknowledging they will sometimes catch harmless requests, though they trigger, on average, in less than 5% of sessions. The company says it is working to reduce these false positives as more capable models arrive in the coming months.
Pricing and autonomy
Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens, which Anthropic says is less than half the price of Claude Mythos Preview. Anthropic says both models can work autonomously for longer than any previous Claude models.
Software engineering and knowledge work
During early testing, Anthropic reports that Stripe said Fable 5 compressed months of engineering into days; in a 50-million-line Ruby codebase, Anthropic says the model performed a codebase-wide migration in a day that would otherwise have taken a whole team over two months by hand. On Cognition's FrontierCode evaluation, which Anthropic describes as testing whether models can pass difficult coding tasks while meeting production-codebase standards, the company says Fable 5 scores highest among frontier models, even at medium effort.
On knowledge work, Anthropic says Fable 5 has the highest score of any model on Hebbia's Finance Benchmark for senior-level reasoning, with substantial gains in document-based reasoning, chart and table interpretation, and problem solving. IMC is reported by Anthropic to have said Fable 5 aced its trading-analysis evaluations nearly across the board.
Vision, memory, and long-context
Anthropic calls Fable 5 the new state-of-the-art model for vision tasks, saying it can extract precise numbers from detailed scientific figures and rebuild a web app's source code from screenshots alone, and that it needed only a minimal, vision-only harness to beat the game Pokémon FireRed, where earlier Claude models had struggled even with additional helper tools. On memory, Anthropic reports that giving Fable 5 persistent file-based memory while playing the game Slay the Spire improved its performance three times more than the same memory gave Claude Opus 4.8, and that Fable reached the game's final act three times more often.
Life sciences: drug design and genomics
Anthropic says its internal protein design experts used Mythos 5 to accelerate aspects of the drug design process by around 10 times, and that in one example, Mythos 5 matched or beat skilled human operators while using protein design and bioinformatics tools with no human assistance, executing the same tasks a scientist normally performs. Nine of 14 protein targets in that study yielded strong candidates Anthropic says it is currently investigating.
Anthropic also reports that Mythos 5 is its first model to consistently produce novel, compelling scientific hypotheses, with the company's scientists preferring Mythos's molecular biology hypotheses over Opus-class models' roughly 80% of the time in blinded comparisons; one such hypothesis, about a novel mechanism for an E. coli protein, was later corroborated by an independent lab. In genomics, Anthropic says Mythos 5 conducted novel research over more than a week of largely autonomous work, assembling single-cell data spanning 138 animal species, and that its trained model outperformed a model published in the journal Science despite being 100 times smaller.
Key facts
- Claude Fable 5 launches for general use; Claude Mythos 5 launches for a small group of trusted cyberdefenders and infrastructure providers.
- Both models are priced at $10 per million input tokens and $50 per million output tokens.
- Safeguards route some Fable 5 queries to Claude Opus 4.8, triggering on average in less than 5% of sessions.
- Anthropic says Stripe found Fable 5 completed a 50-million-line Ruby codebase migration in a day versus over two months by hand.
- Mythos 5 reportedly sped up aspects of drug design by around 10 times and its genomics model was 100 times smaller than one it outperformed.
Check your understanding
What safeguard does Anthropic apply to some Fable 5 queries, and how often does Anthropic say it triggers? Which government program is Claude Mythos 5 initially being deployed through, according to Anthropic?