Anthropic has released Claude Fable 5, a Mythos-class model now in general availability. It is their most capable model to date, designed for sustained multi-day task execution across domains including coding, research, legal analysis, and finance.
The key technical claim is endurance: Fable 5 is built to hold context and progress on complex, long-horizon problems without degradation. Anthropic also states the model ships with safeguards sufficient for broad public release, which signals a specific internal safety threshold has been met, not just capability benchmarks.
The full announcement at anthropic.com details what Mythos-class actually means in practice, what the safeguard architecture looks like, and where the capability ceiling sits. Those specifics are worth reading before drawing conclusions from the headline.
[WATCH ON YOUTUBE →]