Claude Fable 5.1 (Adaptive Reasoning, Max Effort, Default Fallback)

Claude Fable 5.1 is Anthropic's generally available, production-safeguarded deployment of the same underlying weights as Claude Mythos 5.1 (trusted-access only via CVP/LSVP; not cataloged here). It targets demanding reasoning and long-horizon agentic work with text and image input, text output, multilingual and vision support, tool use, adaptive thinking always on (API/Claude Code default effort high; Cowork/claude.ai default medium), a 1M-token context window, and 128K max output on the sync Messages API. Reliable knowledge and training-data cutoffs are June 2026; comparative latency is slower than the rest of the current lineup. First-party pricing matches Fable 5 at $10/$50 per million input/output tokens, with cache reads cut to $0.25 per million tokens (0.025x base input, down from $1 / 0.1x on Fable 5); Anthropic estimates ~25% lower cost on typical workloads and ~45% on highly agentic ones. Self-reported launch benchmarks with production safeguards enabled include Terminal-Bench-Science 0.1 52.6% (SE ±3.5–4.5 pts), Terminal-Bench 4.0 55.8%, GDPval-AA v2 1853 Elo, OSWorld 2.0 77.9% partial / 41.7% strict (August 2026 task release), Humanity's Last Exam 60.9% no tools / 65.0% with tools, AutomationBench 31.4%, and CursorBench 3.2.0 73.4%. Available on the Claude API as `claude-fable-5-1`.