Anthropic has launched Claude Fable 5, its most capable AI model ever. It includes safeguards for general use, while a separate Claude Mythos 5 with relaxed safeguards is available for cyberdefense.
On June 9, 2026, Anthropic launched Claude Fable 5 and Claude Mythos 5. Fable 5 is a Mythos-class model with safeguards for general use. Mythos 5 is the same underlying model with relaxed safeguards in cybersecurity areas, initially deployed through Project Glasswing in collaboration with the US government.
Fable 5 achieves state-of-the-art performance on nearly all benchmarks, including software engineering, knowledge work, vision, and scientific research. Stripe reported that Fable 5 completed a codebase-wide migration of a 50-million-line Ruby codebase in one day. Pricing is $10 per million input tokens and $50 per million output tokens, less than half the price of Claude Mythos Preview.
Fable 5 represents a significant capability leap, highlighting Anthropic's strategy of balancing performance and safety. Safeguards are designed to trigger in less than 5% of sessions, but may occasionally block harmless requests. The limited release of Mythos 5 signals potential for high-capability AI in security applications.
Regarding Anthropic's release of Claude Fable 5 and Mythos 5, some users expressed concerns about the conservative safety measures and the mention of cybersecurity risks. Others criticized previous warnings calling for a halt to AI development as marketing tactics, citing 'recursive self-improvement'. Additionally, some pointed out that the goalposts keep shifting, referencing the 319-page system card and METR's evaluation that 'multi-week R&D automation is impossible'.