Overview
- Anthropic announced Tuesday that Claude Fable 5.1 is broadly available while Claude Mythos 5.1 remains limited to vetted partners with different safety settings.
- Fable 5.1 adds a one million token context window, up to 128k‑token outputs, mid‑conversation effort control, and content provenance tagging to support multi‑step coding and research tasks.
- The company cut cache‑read pricing by 75% to $0.25 per million tokens, which Anthropic says will lower typical workload costs by about 25% and can save up to roughly 45% on highly agentic tasks.
- Anthropic tightened and refined its classifiers to reduce false positives in cyber and biology queries, introduced invisible watermarking with a detection API in private preview, and said Enterprise Frontier Safeguards will let customers keep monitoring data on their own cloud infrastructure.
- Company benchmarks show notable gains on scientific and coding tests and customer trials like Millennium’s bug hunt, and the rollout signals Anthropic’s push to win enterprise use for long‑horizon agent workflows while coordinating restricted access with vetted partners and some government programs.