Anthropic releases Claude Mythos 5 to limited partners and Claude Fable 5 publicly with cybersecurity guardrails

Anthropic released Claude Fable 5 and Claude Mythos 5 on Tuesday. Claude Mythos 5 is available only to a limited set of industry partners and select biology researchers, with collaboration from the US government on the rollout. Claude Fable 5 is publicly released but includes guardrails that block answers to questions about cybersecurity, biology, and chemistry, rerouting those requests to Claude Opus 4.8.

Claude Fable 5 uses the same underlying model as Mythos 5 but with safety restrictions. Anthropic also reroutes requests if it suspects a user is attempting distillation (training a smaller AI model from a larger model's responses). The company conducted over 1,000 hours of red-teaming and found no universal jailbreaks for the model.

Anthropic is pricing Claude Fable 5 and Claude Mythos 5 at $10 per million input tokens and $50 per million output tokens—twice the cost of Anthropic's publicly available AI models but cheaper than Mythos Preview. The limited release reflects concerns that Mythos-level capabilities could be exploited to develop hacking tools, and Anthropic states it has not yet developed fully robust safeguards to prevent misuse.