Shared benchmarks

3
Shared benchmarks for Composer 1.5 and Composer 2
BenchmarkComposer 1.5Composer 2
  • 44.2%
  • 61.3%
Evaluation details for CursorBench 3.0

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Composer 1.5

Record for Composer 1.5 on CursorBench 3.0

Composer 2

Record for Composer 2 on CursorBench 3.0
  • 65.9%
  • 73.7%
Evaluation details for SWE-bench Multilingual

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Composer 1.5

Record for Composer 1.5 on SWE-bench Multilingual

Composer 2

Record for Composer 2 on SWE-bench Multilingual
  • 47.9%
  • 61.7%
Evaluation details for Terminal-Bench 2.0

The recorded benchmark version, metric, and evaluator sets match. Evaluation methodology is not recorded by the registry.

Composer 1.5

Record for Composer 1.5 on Terminal-Bench 2.0

Composer 2

Record for Composer 2 on Terminal-Bench 2.0