Details
- Cursor announced that the Claude Fable 5 model is now available directly inside the Cursor AI code editor.
- The company reports that Fable 5 achieved a 72.9% score on CursorBench, setting a new state-of-the-art result on its internal autonomous coding benchmark.
- CursorBench rankings show Fable 5 Max in first place at 72.9%, ahead of other Fable 5 tiers and previous top-performing models like Opus 4.7 and GPT-5.5 variants.
- Cursor highlights that Fable 5’s score is roughly 8 percentage points higher than the prior best model on CursorBench, indicating a significant step up in end-to-end coding task performance.
- Users can compare Claude Fable 5 against other leading models via Cursor’s published benchmark leaderboard, which includes performance, cost, and token usage metrics across multiple SKUs.
- The integration positions Fable 5 as a first-class option in Cursor alongside models from Anthropic, OpenAI, and Cursor’s own Composer line, giving developers another high-end choice for complex coding workflows.
Impact
By bringing the top-scoring CursorBench model directly into the editor, Cursor strengthens its positioning as a neutral, performance-driven hub for frontier coding models. This move raises competitive pressure on both in-house models like Composer and external rivals from OpenAI and Anthropic, and may accelerate a shift toward benchmark-driven model selection in day-to-day software development.
