Anthropic opens Claude Fable 5 to public; Mythos 5 remains restricted

Anthropic has made Claude Fable 5 available to the public while keeping Claude Mythos 5 in a restricted program. Fable 5 is offered as a Mythos-class model with added safety controls. Mythos 5 is limited to Project Glasswing partners and a small number of approved biology researchers.

Fable 5 is available immediately on Anthropic’s Pro, Max, Team and Enterprise plans through June 22, 2026. After June 22, use may require credits until Anthropic expands capacity. The company set Fable’s price at $10 per million input tokens and $50 per million output tokens.

Access to Claude Mythos 5 is confined to Project Glasswing participants, including technology and security partners such as AWS, Microsoft, Apple and CrowdStrike, and to a select group of biology researchers. Business customers using Mythos-class models must accept a mandatory 30-day data retention window so Anthropic can monitor for safety issues.

Anthropic described both Fable 5 and Mythos 5 as high-performance models for coding, reasoning, vision, scientific research and long-context tasks. The company reported that Mythos 5 delivered speedups in drug-design workflows and that researchers preferred novel scientific hypotheses produced by the model about 80% of the time. Anthropic also reported a week-long autonomous genomics experiment in which Mythos 5 outperformed a recent published model while being roughly 100 times smaller.

Anthropic cited a customer example in which Fable 5 accelerated engineering work on a large Ruby codebase. Stripe reported that what would have taken a team more than two months was compressed into days when using Fable 5.

The company built layered safeguards into the public rollout. Requests flagged as involving sensitive cybersecurity, biological, chemical or model-distillation content are automatically routed to Claude Opus 4.8. Anthropic said those fallback safeguards trigger in fewer than 5% of sessions.

The Mythos-class architecture first appeared in a restricted preview through Project Glasswing in April 2026. During that preview, the model demonstrated an ability to identify and chain zero-day exploits across major operating systems and browsers, which prompted additional safety controls for the public release.

An analyst at Bull Theory noted, “Claude Fable will be roughly twice as expensive as today’s most advanced Claude Opus models.” Anthropic has also disclosed recent fundraising and a confidential IPO filing in materials describing its strategy to serve infrastructure and research customers.

Anthropic said it will expand access and capacity over time and that it has routinized safety checks and fallback handling to manage risk while offering the new model to a broader set of users.