CursorBench 4.1
We evaluate agents on ambiguous, multi-file tasks from real Cursor sessions. Higher scores are better.
More about CursorBench| Model | |||||
|---|---|---|---|---|---|
| 1 | Gemini 4 Neon Extra High | 65.4% | $10.67 | 136,273 | 73 |
| 2 | Gemini 4 Neon High | 63.7% | $6.54 | 88,596 | 56 |
| 3 | Gemini 4 Carbon High | 61.5% | $8.52 | 114,518 | 97 |
| 4 | Gemini 4 Hydrogen Extra High | 61.4% | $8.09 | 205,981 | 113 |
| 5 | Fable 5.2 Max | 61.3% | $48.11 | 541,003 | 295 |
| 6 | Gemini 4 Carbon Medium | 60.4% | $7.22 | 103,189 | 83 |
| 7 | Fable 5.2 Extra High | 58.8% | $17.84 | 194,918 | 122 |
| 8 | Opus 5.5 Max | 58.1% | $11.88 | 213,070 | 144 |
| 9 | Gemini 4 Neon Medium | 57.7% | $3.07 | 47,059 | 41 |
| 10 | Gemini 4 Hydrogen High | 57.3% | $4.57 | 113,807 | 77 |
| 11 | Opus 5.5 Extra High | 56.7% | $6.54 | 104,901 | 95 |
| 12 | Fable 5.2 High | 56.5% | $10.51 | 108,213 | 77 |
| 13 | Gemini 4 Argon High | 56.5% | $9.14 | 135,038 | 118 |
| 14 | Gemini 4 Argon Medium | 55.8% | $7.64 | 128,231 | 101 |
| 15 | Gemini 4 Carbon Low | 55.0% | $8.64 | 48,526 | 127 |
| 16 | Opus 5.5 High | 54.5% | $3.87 | 56,162 | 64 |
| 17 | Sonnet 5.5 Max | 54.4% | $8.63 | 342,273 | 199 |
| 18 | Fable 5.2 Medium | 54.2% | $5.01 | 47,678 | 40 |
| 19 | Fable 5.2 Low | 53.4% | $4.26 | 40,384 | 34 |
| 20 | Opus 5.5 Medium | 50.0% | $2.85 | 39,736 | 52 |
| 21 | Sonnet 5.5 Extra High | 49.9% | $3.13 | 112,299 | 87 |
| 22 | Grok 4.7 Extra High | 48.6% | $10.19 | 90,268 | 110 |
| 23 | Haiku 5.5 Max | 48.4% | $1.14 | 344,890 | 166 |
| 24 | Gemini 4 Argon Low | 47.9% | $7.96 | 40,541 | 121 |
| 25 | Gemini 4 Hydrogen Medium | 46.4% | $1.28 | 43,606 | 45 |
| 26 | Grok 4.7 High | 46.0% | $6.88 | 65,850 | 84 |
| 27 | Gemini 4 Neon Low | 45.5% | $1.26 | 19,580 | 26 |
| 28 | Haiku 5.5 Extra High | 45.2% | $0.56 | 149,314 | 94 |
| 29 | Grok 4.6 Extra High | 44.3% | $5.26 | 58,554 | 62 |
| 30 | Sonnet 5.5 High | 43.8% | $1.41 | 43,088 | 48 |
| 31 | Opus 5.5 Low | 42.7% | $1.43 | 19,497 | 32 |
| 32 | Grok 4.7 Medium | 42.5% | $5.52 | 41,173 | 70 |
| 33 | Muse Spark 1.3 Max | 41.9% | $2.89 | 56,572 | 113 |
| 34 | Grok 4.6 High | 41.3% | $4.28 | 46,330 | 52 |
| 35 | GPT-5.6 Sol Max | 40.9% | $7.60 | 40,466 | 100 |
| 36 | Grok 4.6 Medium | 38.9% | $2.59 | 26,374 | 40 |
| 37 | Haiku 5.5 High | 37.5% | $0.31 | 77,196 | 61 |
| 38 | GPT-5.6 Sol Extra High | 37.4% | $4.32 | 23,921 | 58 |
| 39 | Muse Spark 1.3 Extra High | 34.6% | $2.33 | 47,142 | 96 |
| 40 | Sonnet 5.5 Medium | 34.3% | $0.63 | 19,144 | 26 |
| 41 | Haiku 5.5 Medium | 34.2% | $0.17 | 43,441 | 41 |
| 42 | GPT-5.6 Sol High | 33.2% | $3.00 | 16,319 | 44 |
| 43 | Grok 4.6 Low | 33.1% | $1.76 | 16,863 | 33 |
| 44 | Muse Spark 1.3 High | 32.1% | $1.97 | 36,562 | 85 |
| 45 | Sonnet 5.5 Low | 31.8% | $0.44 | 13,798 | 21 |
| 46 | Grok 4.7 Low | 31.5% | $2.73 | 14,493 | 42 |
| 47 | Muse Spark 1.3 Medium | 31.3% | $1.75 | 32,034 | 77 |
| 48 | GPT-5.6 Sol Medium | 28.7% | $1.97 | 10,554 | 34 |
| 49 | Haiku 5.5 Low | 27.8% | $0.08 | 23,673 | 27 |
| 50 | Muse Spark 1.3 Low | 27.5% | $1.13 | 20,673 | 58 |
| 51 | Gemini 4 Hydrogen Low | 23.9% | $0.28 | 10,497 | 22 |
| 52 | GPT-5.6 Sol Low | 23.0% | $0.98 | 5,206 | 22 |
| 53 | Muse Spark 1.3 Minimal | 21.3% | $0.69 | 12,789 | 42 |
Changelog
Tasks
- CursorBench 4.1
- Modified instruction following problems.
- Increased agent time limit.
Reporting
- Updated Sonnet 5.5 results to account for adjusted pricing.
Tasks
- CursorBench 4.0
- Introduced new long-horizon problems focused on edit, refactor, investigation, intent understanding, managing jobs, and design adherence.
Reporting
- Updated Sonnet 5 results to account for adjusted pricing.
Reporting
- Updated GPT-5.6 Terra and Luna results to account for adjusted pricing.
Reporting
- Updated GPT-5.6 Sol, Terra, and Luna results to account for cache write costs.
Tasks
- CursorBench 3.2
- Introduced instruction following and advanced tool use problems.
Tasks
- CursorBench 3.1
- Introduced problems focused on codebase understanding, bugfinding, planning, and code review.
- Improved grading criteria for some edit tasks.
Tasks
- CursorBench 3.0
- Initial set of tasks focused on edit, refactor, and bugfix problems.
Avg cost / task is computed by applying each model's published per-million-token pricing (input, cache read, cache write, and output) to the tokens it used on each task. Results are subject to variance; small differences in scores may not be statistically meaningful.