Fable 5 is Anthropic's public release of Mythos, the frontier model the company previewed in April.
Anthropic first announced Mythos as a limited preview in April, pitching it as a highly capable frontier model. The public version, now called Fable 5, is available to all. Anthropic says it is "safe for general use" and claims state-of-the-art standing on nearly all tested benchmarks.
That phrase "safe for general use" is doing real work. It implies the April preview was not ready for the public, which means Anthropic ran the model through meaningful safety and evaluation work in the months between then and now. The "nearly all benchmarks" qualifier deserves equal scrutiny: benchmark selection is a lab's own choice, and "nearly" tends to paper over the results they'd rather not highlight.
The preview-to-public pipeline has become standard practice at frontier AI labs. Show capabilities to a limited audience, refine what's risky or rough, then open the doors. Fable 5 arrived on that familiar schedule. What matters now is whether independent evaluators land in the same place Anthropic did.