Claude Fable 5 “predicting” a solar eclipse.
Credit: Anthropic
Anthropic has released a new version of its Mythos-class AI model—now called Claude Fable 5—to the public. The company says Fable 5 uses the same Mythos base model it previously restricted for safety reasons.
The original Claude Mythos 5 model is only available to approved partners and government programs that work on cybersecurity and other important infrastructure. In these scenarios, Anthropic relaxes some of the model’s most rigorous guardrails to let experts use it in probe-and-defend systems. For general users, however, Fable 5 will act as the public face of Mythos‑class capability.
Fable 5 uses strict rules to block responses to sensitive topics such as cybersecurity and biological threats. Anthropic says this lets the company keep most of Mythos’ capabilities accessible for everyday work, from coding and research to analysis and content creation, without putting the public at risk.
Fable 5 sits above Claude Opus 4.8 in Anthropic’s model lineup and is now the most capable model available to regular users. Early benchmarks indicate that Fable 5 outperforms Opus 4.8 by a noticeable margin across coding, reasoning, scientific tasks, and more. Anthropic notes that Fable 5 can work on complex projects for days, making it useful in software engineering, finance, legal work, or anything that requires a longer timeframe.
The company frames Claude Fable 5 as proof that it can bring high‑end AI models to regular users without exposing the riskiest behaviors that raised concerns around the original Mythos release.
