Claude Fable 5.1 and Claude Mythos 5.1

Anthropic has introduced Claude Fable 5.1 and Claude Mythos 5.1, described as the world's most advanced models for coding and knowledge work. The two are the same underlying model but with different safeguard configurations: Fable 5.1 is generally available, while Mythos 5.1 is restricted to trusted access programs aimed at cybersecurity and life sciences.

On cost, Fable 5.1 is estimated to be about 25% cheaper than its predecessor Fable 5 for typical token-billed workloads, with savings up to ~45% for highly agentic work, driven by reduced pricing on cache reads. Anthropic also announced Enterprise Frontier Safeguards (EFS), a data-retention system that keeps data in customer-controlled cloud infrastructure, offering complete privacy while still preventing adversarial use; EFS will roll out to enterprises in phases starting fall, and until then eligible customers can use Fable 5.1 with zero data retention. Safeguard improvements cut false positives by 60% in cybersecurity, and Fable 5.1 can now discover software vulnerabilities (though not develop exploits). In biology, an access program developed with the US government will soon open for scientists to use Claude Mythos 5.1’s advanced capabilities.

On performance, Fable 5.1 sets a new standard for coding, knowledge work, and long-running problems. In the Terminal-Bench-Science 0.1 benchmark, accuracy versus cost plots show Fable 5.1 outperforming Fable 5, and at Low or Medium effort it matches or beats Fable 5 at lower cost (Fable 5.1 defaults to High effort in Claude Code and Medium in Claude Cowork/Claude.ai). The reported standard error is ±3.5–4.5 points per model; public leaderboard scores (3 trials/task, Claude Code harness) list Claude Opus 5 at 30.0% and Claude Fable 5 at 21.4%, with Anthropic’s setup reproducing them at 29.0% and 24.7%, respectively, both within noise.