BenchLeader
CursorAuto-detected

Composer 2.5

Not yet ranked: too few independent results so far. Released 18 May 2026.

Blended price
Output speed
First answer
Context
How it scores by categoryDashed line = average model (50). One step of 15 = one standard deviation.
  1. Coding40
  2. Agents & tools47

Benchmark results

One column per reasoning effort. Rank is among every configuration of every model on that benchmark. Hover a score for the run it came from.

Coding

BenchmarkdefaultSource
FrontierCode25.6%#21Cognition
SWE-bench (Vals)not in index79.6%#25Vals AI

Agents & tools

BenchmarkdefaultSource
Terminal-Bench 2.1 (Vals)58.4%#32Vals AI

See also

Data as of 9 Sept 2026. Compare with another model.