Anthropic launches Claude Fable 5, its first 'Mythos-class' AI model with new safeguards

Anthropic has unveiled Claude Fable 5, the first widely available model from its previously unreleased Mythos class. The company had earlier described the Mythos family as too dangerous for public release due to its advanced cybersecurity capabilities. Now, Anthropic says new safeguards make the release possible, blocking responses in specific high-risk areas such as cybersecurity and biology, and falling back to the 'honest' Claude Opus 4.8 when necessary. In testing, 95% of sessions ran entirely on Fable without requiring fallback.

Fable 5 shows exceptional performance across software engineering, knowledge work, and vision tasks, with its advantage expanding as tasks grow longer and more complex. Anthropic is also releasing Claude Mythos 5, which is the same underlying model but with safeguards lifted in some areas, initially available to organizations in the private Project Glasswing initiative. Pricing is $10 per million input tokens and $50 per million output tokens—double Opus 4.8 but half the cost of Mythos Preview. Anthropic plans to expand access to Mythos 5 over time through a systematic trusted-access program.