Anthropic launches Claude Fable 5.2, limits Mythos 5.2 access

Anthropic launched Claude Fable 5.2 and Claude Mythos 5.2 on November 9. Fable is generally available, while Mythos is limited to eligible cybersecurity and life-sciences organizations admitted to Anthropic’s verification programs. The company said both models are aimed at demanding coding, reasoning and technical work.

According to Anthropic, the two products use the same underlying model. Fable adds cybersecurity and biology safeguards for broad access; Mythos removes those extra restrictions for approved organizations. The company’s published materials did not fully specify the programs’ acceptance criteria, review timing or geographic eligibility.

Standard pricing for both models is $8 per million input tokens and $40 per million output tokens. An optional fast mode costs $16 per million input tokens and $80 per million output tokens. Anthropic said fast mode can increase token-generation speed by as much as 2.5 times, while the regular models generate output up to 30% faster than their 5.1 predecessors. Neither claim has been independently verified.

The Fable 5.2 model documentation lists the API identifier as claude-fable-5-2, a 1-million-token context window and a maximum output of 128,000 tokens. Anthropic lists Fable as active through the Claude API, Amazon Bedrock, Claude Platform on AWS, Google-managed and Anthropic-managed Claude Models on Google Cloud, and Microsoft Foundry. Google Cloud separately lists Fable 5.2 as generally available with a November 9 release date.

Anthropic said fast mode is available through Claude Code, the Claude Platform, both Google-managed and Anthropic-managed Claude Models on Google Cloud, and Anthropic-hosted Microsoft Foundry. Mythos fast mode remains restricted to organizations in the verification programs.

In Anthropic’s evaluations, Fable 5.2 scored 49.1% on FrontierCode 2.0 at Xhigh effort, 69.9% on FrontierSWE v2 and 83.5 on BenchCAD 2.1 at Max effort. The results are company evaluations, not independent findings.

Anthropic also said prerelease users solved two open black-hole-physics problems, designed three working electronic devices and formalized Fermat’s Last Theorem in Lean in 30 hours. Those scientific-achievement claims have not been independently verified.

On safety, Anthropic said its prerelease testing indicated a low risk of high-stakes misalignment and that Mythos performed as well as or better than recent Claude models on most alignment measures. The company also said Fable’s cyber and biology classifiers produced about 55% fewer blocks than Fable 5.1 in early testing while retaining safeguards it described as equally robust. Those safety results are Anthropic’s assessments and have not been independently replicated.