Claude Fable 5.1 is the latest entry in Anthropic's Fable line, positioned by the company as its most capable model for coding and professional knowledge work, with research capabilities that the vendor describes as an early glimpse of how AI will contribute to scientific progress. It belongs to the same fifth-generation family that introduced longer-running, asynchronous workflows, and the 5.1 update sharpens that focus: Anthropic highlights strengths in agentic coding, sustained multi-step problem solving, and tasks that span days rather than minutes. The model is also offered in a restricted sibling, Mythos 5.1, which shares the underlying weights but ships under a tighter safeguard regime aimed at vetted cybersecurity and life sciences users, while Fable 5.1 itself remains the broadly available variant for general enterprise and developer use.
In benchmark reporting shared at launch, Fable 5.1 posts leading scores across the evaluations Anthropic selected, including a 52.6% result on Terminal-Bench-Science 0.1, more than double the Fable 5 baseline, and a 55.8% score on Terminal-Bench 4.0 against 42.0% for the prior generation, with the comparison table covering Fable 5, Opus 5, and GPT-5.6 Sol. Alongside the capability gains, the release trims cache-read pricing by roughly 75%, a meaningful change for teams that reuse large prompts and depend on prompt caching for cost control. With a generous working context window, native handling of text plus image and PDF inputs, and tool calling plus structured output for agent pipelines, Fable 5.1 fits well for long-running coding sessions, research assistants, document-heavy workflows, and enterprise applications where sustained reasoning matters more than short, single-turn exchanges.
