Anthropic launches Claude Fable 5 with a gated Mythos 5 for cyber use

Why it matters

Anthropic is pushing a two-tier release strategy for frontier models: general access with automated routing away from risky prompts, and a less restricted channel for vetted cyber users.

Anthropic launched Claude Fable 5 today, alongside a more restricted Claude Mythos 5 for cyberdefenders and infrastructure providers, in a blog post that frames the release around capability and containment.

Anthropic says Fable 5 is the most capable model it has made generally available and describes it as state-of-the-art on nearly all tested benchmarks, including software engineering, knowledge work, vision and scientific research. Those are Anthropic's benchmark claims, not independently verified results. The most important product detail is the safety architecture: Anthropic says requests on some topics will instead be answered by Claude Opus 4.8, with the conservative filter triggering in less than 5% of sessions on average.

Mythos 5 uses the same underlying model, according to Anthropic, but lifts some safeguards for a narrower audience. Anthropic says Mythos 5 will initially roll out through Project Glasswing, its collaboration with the US government, as an upgrade to Claude Mythos Preview. Anthropic also says it intends to expand access through a broader trusted access program.

The pricing is explicit: both models cost $10 per million input tokens and $50 per million output tokens, which Anthropic says is less than half the price of Claude Mythos Preview. In early testing cited by Anthropic, Stripe reported that Fable 5 completed a 50-million-line Ruby migration in a day, while Fable 5 scored highest among frontier models on Cognition's FrontierCode evaluation. Anthropic also cited benchmarks from Hebbia and IMC, plus a vision-only run in which Claude played Pokemon FireRed.