HomePrice per Intelligence Point and Cost per Task
CursorBench 3.2: Agentic Coding Score × Cost per Task (Effort Sweep, Snapshot Sep 2 2026)
CursorBench 3.2: Agentic Coding Score × Cost per Task (Effort Sweep, Snapshot Sep 2 2026)#
On the real-world coding benchmark Fable 5.1 took first place while costing about half as much per task as its predecessor.Draft, pending review
How to read this chart
Cursor's official benchmark, built from real multi-file, loosely specified tasks. X is actual cost per task (log), y is score; dots of one color on one line are the same model swept across effort settings from low to high.
Source: Cursor public benchmarks; FinSight compilation · Updated 2026-09-04
↩ This chart lives on the paged3-intel-price-mon-06