Benchmark Almanac

Cursor

4 models 4 model cards read

What its model cards measure

Tall ticks are Cursor model cards that list the benchmark in their headline table. Gray dots are cards that leave it out. Rows are the benchmarks the lab listed most, at least twice each, in the order it started using them.

Cursor Bench (2025 internal version) Terminal-Bench 2.0 CursorBench 3.1 SWE-bench Multilingual

Model cards

DateModelDocumentBenchmarks
18 May 2026 Composer 2.5 Introducing Composer 2.5 (Cursor blog) 3
19 Mar 2026 Composer 2 Introducing Composer 2 (Cursor blog) 3
9 Feb 2026 Composer 1.5 Introducing Composer 1.5 (Cursor blog) 2
29 Oct 2025 Composer 1 Composer: Building a fast frontier model with RL (Cursor blog) 1

Every Cursor model