Anthropic launched Claude Fable 5, a Mythos‑class model safe for general use, surpassing all previously available models on benchmarks, excelling in software engineering, knowledge work, vision, and scientific research. Safety layer redirects about 5% of queries to Claude Opus 4.8. Pricing is $10 per million tokens for input and $50 for output, keeping the model affordable for both startups and enterprises.

For a limited set of cyber defenders and infrastructure providers Anthropic released Claude Mythos 5, the same engine as Fable 5 but with fewer safeguards. Deployed via Project Glasswing in partnership with the US government, Mythos 5 already accelerated code‑migration tasks, helped protein‑design teams achieve ten‑fold speedups, and has already been used in critical infrastructure audits, generating novel hypotheses that independent labs later validated.

Independent tests show Fable 5 compressing months of engineering into days; Stripe reported a 50‑million‑line Ruby migration completed in a single day, while finance benchmarks place it atop senior‑level reasoning scores. Anthropic’s alignment assessment finds misalignment rates comparable to Opus 4.8, and demonstrates that large‑scale autonomous reasoning can be trusted in production pipelines.