Anthropic Releases Claude Fable 5.1 and Claude Mythos 5.1
Fable 5.1 doubles the scientific benchmark score of Fable 5 and cuts agent-task costs by up to 45%, directly addressing the false-detection and weak agent performance complaints that followed the previous release.
Key Facts
- Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026.
- Claude Fable 5.1 scored 52.6% on the Terminal-Bench-Science 0.1 benchmark, more than double Claude Fable 5's 24.7% and above Claude Opus 5's 29.0%.
- The cache read price for Claude Fable 5.1 fell 75%, from $1.00 to $0.25 per million tokens, while input and output pricing stayed at $10 and $50 per million tokens.
- Cybersecurity false detections in Claude Code fell about 60% and elementary biology and medical false detections fell 85% compared with the Fable 5 launch safeguards.
- Claude Mythos 5.1 is offered only through a review program for cybersecurity and life sciences specialists, currently limited to some US organizations.
Reporting from 2 sources: GameBusiness.jp, GIGAZINE.
Anthropic released Claude Fable 5.1 and Claude Mythos 5.1 on September 1, 2026. Fable 5.1 is the general release with security safeguards, while Mythos 5.1 removes those limits and is available only to approved organizations. Fable 5.1 cuts costs by 25% for general workloads and up to 45% for complex agent tasks. The cache read price dropped 75%, from $1.00 to $0.25 per million tokens, while input and output pricing stays at $10 and $50 per million tokens. On the Terminal-Bench-Science 0.1 benchmark, Fable 5.1 scored 52.6%, more than double Fable 5's 24.7% and well above Opus 5's 29.0%. Safety changes cut cybersecurity false detections by about 60% and elementary biology and medical false detections by 85%. Fable 5.1 is available immediately through the API, Amazon Web Services, Google Cloud, and Microsoft Azure, and on Pro, Max, Team, and Enterprise subscription plans.
Anthropic calls Fable 5.1 its most advanced coding and knowledge work model. On the Terminal-Bench 4.0 agentic coding benchmark, Fable 5.1 scored 55.8%, about 3.5 points above Opus 5's 52.3%. On Humanity's Last Exam with tool use, the gap narrows to 65.0% against 63.6%. These scores come from Anthropic's maximum effort setting; the default in Claude Code is High and in Claude Cowork and Claude.ai it is Medium, so typical use will not reproduce them exactly.
The two models are equivalent in capability. Fable 5.1 carries security safeguards that reject dangerous tasks or route them to the Opus line, while Mythos 5.1 removes those limits and is offered only through a review program for cybersecurity and life sciences specialists, currently limited to some US organizations. Anthropic now permits Fable 5.1 for software vulnerability discovery, but exploit development, penetration testing, and binary analysis vulnerability scanning still route to Opus. The Claude Security vulnerability scanning service was updated to run on Mythos 5.1 the same day.
Anthropic also reset subscription usage limits. Fable 5.1 is available within Max and equivalent enterprise plans up to half of the usage allowance. The prompt guide and a system card covering both models were published alongside the release.
Synthesized by Yomimono from the 2 cited sources below, including Japanese-language reporting where cited, then editorially reviewed before publishing.