Claude Fable 5.1 launches on Cursor with 73.4% benchmark score
dean_rie · x · 2026-09-02
Claude Fable 5.1 is now available in Cursor. It achieved a score of 73.4% on CursorBench 3.2, making it the most capable model the platform has run. The team found it particularly skilled at self-verification, enabling it to handle difficult coding tasks from start to finish.
More from coding & agent
- Perplexity launches hybrid local/cloud agent architecture with privacy switch — AccBalanced · 2026-09-02
- Mitsuhiko: Pi's hookable system prompt makes long-term prefix stability the hard problem — mitsuhiko · 2026-09-02
- Anthropic's Model Switch Limitation Sparks Stability Concerns — mitsuhiko · 2026-09-02
- Mitsuhiko on evaluating EffectTS: "can I read and understand the code" — mitsuhiko · 2026-09-02
- Why Pi Codebase Hesitates on Adopting EffectTS — mitsuhiko · 2026-09-02
- Guide: Integrating ChatGPT 5.6 Pro into Codex workflows — OpenAIDevs · 2026-09-02