Anthropic Launches Claude Fable 5.1, Topping FrontierSWE v2

Anthropic officially released Claude Fable 5.1, achieving a record 56.3% score on the FrontierSWE v2 software engineering benchmark. The model nearly doubles the performance of GPT-5.6 Sol on multi-file codebase refactoring and autonomous architectural debugging. The release features optimized agentic planning tokens and deterministic tool calling for complex software production environments.