Cursor Adds Claude Opus 5.5

The model leads CursorBench at 57.8% and costs 40% less per task than Opus 5.

Cursor users can run tasks on Claude Opus 5.5 with higher benchmark accuracy and lower per-task costs.