Anthropic releases Claude Fable 5 and Mythos 5 with major gains in coding and science

Details

Anthropic released two fifth-generation models today. Claude Fable 5 is the general-availability release with standard safety guardrails; Claude Mythos 5 shares the same base but with relaxed restrictions for areas like cybersecurity, and is currently limited to a small group of selected partners. The benchmark gains are substantial. On SWE-Bench Pro (real-world GitHub software engineering tasks), Fable 5 scores 80.3% — versus Claude Opus 4.8 at 69.2%, GPT 5.5 at 58.6%, and Gemini 3.1 Pro at 54.2%. On Cognition's FrontierCode benchmark (production-standard coding tasks), Fable 5 scores 29.3% against Opus 4.8's 13.4% and GPT 5.5's 5.7%. Real-world evidence backs the benchmarks. Stripe ran Fable 5 on a 50-million-line Ruby codebase and completed a migration in one day that would have taken a full engineering team two months. Anthropic also claims improved token efficiency at medium effort compared to prior Claude generations. Mythos 5 is the more restricted sibling: it autonomously designed drug candidates and operated largely independently in genomics research, but its offensive cybersecurity capabilities prompted Anthropic to gate access exclusively to select partners. Both models are priced at $10 per million input tokens — roughly 2x Claude Opus 4.8 — with token efficiency in production still to be fully characterized.