Composer 2.5: benchmark scores, price and ranking
Composer 2.5 from Cursor ranks #497 of 743 LLMs on the LLMs Tiger aggregate benchmark score (46.6), based on 2 benchmark results captured up to 2026-10-13.
- Organisation
- Cursor
- Type
- Frontier model
- Open weights
- No
- Released
- 2026-05-18 (source: epoch)
- Aggregate rank
- #497 of 743 (score 46.6)
Across the 2 benchmarks where at least ten models are scored, Composer 2.5 is in the top 10% on 0 and in the top half on 0.
Composer 2.5 benchmark results
| Benchmark | Score | Rank | Source | Captured |
|---|---|---|---|---|
| FrontierCode | 25.6% | #10 of 15 | Epoch AI Benchmarking Hub | 2026-10-13 |
| CursorBench | 27.7% | #15 of 15 | Epoch AI Benchmarking Hub | 2026-10-13 |
About Composer 2.5
Composer 2.5 is a frontier model from Cursor. It was released on 2026-05-18, according to Epoch AI. No API price is listed for it yet. We hold 2 benchmark results for it across 1 area, and none is in the top tenth of its board.
Where Composer 2.5 stands
On our aggregate score, Composer 2.5 is #497 of 743 ranked models, in the lower half. Its score is 46.6; the scale is centred on 50, so it sits 3.4 points below the typical model. That beats 33% of the ranked field.
Composer 2.5, benchmark by benchmark
Composer 2.5 lands below the middle on FrontierCode: 25.6%, #10 of 15, with 60% of the field scoring higher. Claude Opus 5 (max) leads at 53.4%; the closest result above is Kimi K2.7 Code at 30.1%. It covers hard programming tasks. The board median is 31.8%, so Composer 2.5 is 6.2 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
CursorBench is a weak spot. Composer 2.5 scores 27.7%, #15 of 15, behind 93% of the models measured. The leader, Claude Opus 5.5 (max), reaches 57.8%. It covers real-world coding-agent tasks. The board median is 41.6%, so Composer 2.5 is 13.9 points below it. Source: Epoch AI Benchmarking Hub, captured 2026-10-13.
Strengths and weaknesses
Software engineering: weak, ahead of 17% of the field on average across 2 benchmarks. Its best board here is FrontierCode (#10 of 15) and its weakest is CursorBench (#15 of 15).
Reading these numbers
The 2 scores on this page come from 1 source: Epoch AI Benchmarking Hub (2). No official leaderboard supplied any of them, so treat them as third-party measurements. With 2 results the aggregate rank can move a lot when one new score arrives.
Models ranked nearby
- #494 mimo-v2.5 (46.7)
- #495 Claude Sonnet 3.5 (46.7)
- #496 MCTS Refine 7B (46.7)
- #498 gemini-2.0-flash-lite (46.6)
- #499 Mellum2-12B-A2.5B-Thinking-SFT (46.6)
- #500 EXAONE-3.5-7.8B-Instruct (46.6)
Compare it on the live leaderboard → · How the aggregate score works