Anthropic releases Claude Fable 5, its first Mythos-class model for the public, and restricted Mythos 5

Anthropic/Claude models

Anthropic released Claude Fable 5 to all users and Claude Mythos 5 to a restricted group of cybersecurity partners. Fable 5 is the first Mythos-class model safe for general use, with automatic fallbacks to Claude Opus 4.8 when detecting requests about cybersecurity, biology with dual-use concerns, or model distillation attempts. Over 95% of Fable sessions require no fallback.

• Fable 5 priced at $10/M input and $50/M output tokens; Mythos 5 restricted to Project Glasswing partners initially
• Both models excel at software engineering, knowledge work, vision processing, drug design, and scientific research across millions of tokens
• Safety classifiers automatically redirect sensitive queries to less-capable Claude Opus 4.8 to mitigate risks from frontier-class capabilities
• Mythos 5 planned for expansion to biomedical researchers after initial cybersecurity rollout