Anthropic launches Fable 5.1 broadly while restricting Mythos 5.1

Signals Inbox·September 2, 2026·Frontier AI

Anthropic released Fable 5.1 broadly while keeping the same underlying model, Mythos 5.1, behind vetting for advanced cyber and biology work. That split is important: frontier labs are experimenting with capability-tiered access, where the model a user gets depends partly on whether safeguards are allowed to step in.

Q1What did Anthropic officially release?

Anthropic's official Mythos page says Claude Mythos 5.1 is its newest high-capability model for cybersecurity and biology, while access remains limited to vetted organizations.

Q2How is Fable 5.1 related?

Anthropic says Fable 5.1 and Mythos 5.1 use the same underlying model. Fable adds safeguards that intervene on risky cyber and biology requests, allowing a version of the capability to be offered much more broadly.

Q3Why restrict Mythos?

Anthropic says the model's abilities in cybersecurity and biology could be misused for attacks or dangerous biological work. Vetted programs give legitimate researchers more capability while preserving tighter controls on who can access it.

Q4How expensive is it?

Anthropic lists Mythos 5.1 starting at $10 per million input tokens and $50 per million output tokens. That is premium pricing aimed at high-value research and security workloads rather than cheap commodity inference.

Q5What is the bigger experiment?

AI access may become segmented by capability and trust. Instead of one public model for everyone, labs can expose more dangerous capabilities only to users who accept verification, monitoring and stricter terms.