Anthropic launches Claude Fable 5, its first public Mythos-class model

Anthropic Releases Claude Fable 5, Its First Public Mythos-Class AI

Anthropic on Tuesday launched Claude Fable 5, making a Mythos-class AI model publicly available for the first time — a move the company had promised since restricting its most powerful system to a handful of cyber defenders in April.

The model shares its underlying architecture with Claude Mythos 5, which remains limited to organizations approved through Anthropic’s Project Glasswing program. The difference lies in the guardrails: when Fable 5’s classifiers detect queries related to cybersecurity, biology and chemistry, or model distillation, the system automatically routes the response to Claude Opus 4.8 instead. Anthropic says more than 95% of sessions involve no such fallback.anthropic

What Fable 5 Can Do

Anthropic describes Fable 5 as exceeding any model it has previously made generally available, with particular strengths in software engineering, knowledge work, and vision. Early testing partners have reported striking results: Stripe said the model compressed months of engineering work into days, performing a codebase-wide migration in a single day across 50 million lines of Ruby code that would have taken a team over two months.anthropic

The model also beat Pokémon FireRed using only a minimal vision-based harness — a task that had stumped previous Claude models even with additional scaffolding tools. On financial reasoning benchmarks, Fable 5 achieved the highest score of any model tested.anthropic

Safety Trade-offs and Data Retention

The release comes with a mandatory 30-day data retention policy on all traffic for Mythos-class models, applicable to both first- and third-party surfaces. Anthropic says the retained data will not be used for training and will be deleted after the 30-day window in most cases.anthropic

Anthropic acknowledged the safeguards are “stricter than would be ideal,” noting that benign requests sometimes trigger the classifiers. An external bug bounty produced no universal jailbreaks in over 1,000 hours of testing, though the UK AI Safety Institute made progress toward one during initial evaluations.anthropic

Pricing and Access

Both Fable 5 and Mythos 5 are priced at $10 per million input tokens and $50 per million output tokens. For subscription users on Pro, Max, Team, and Enterprise plans, Fable 5 is available at no extra cost through June 22, after which usage credits will be required.anthropic

Claude Mythos 5 — the same model with cyber safeguards lifted — is available immediately to existing Glasswing partners, with Anthropic planning a broader trusted access program in consultation with the US government. A separate biology-focused trusted access track will enroll researchers needing the model’s chemistry and biology capabilities without restrictions.anthropic

Tech journalist Alex Heath first reported the imminent release on Sunday, writing that Anthropic was “planning to release a public version of Mythos” with “substantial guardrails.”sources