Calibration
CursorBench 4.0 → SWE-bench Pro
SWE-bench Pro is estimated from CursorBench 4.0 with a Hill curve fitted on 9 models measured on both: y = 0.5849 + (1.0973 − 0.5849)·x^6.00 / (0.54212^6.00 + x^6.00), R² = 0.93, cross-validated error 4.2 pp. It is used for 8 estimates.
| Estimated model | CursorBench 4.0 | SWE-bench Pro | Source |
|---|---|---|---|
| Gemini 4 Carbon | 60.4% | 92.1% | estimated ± 4.2 pp, medium confidence |
| Gemini 4 Hydrogen | 54.3% | 84.2% | estimated ± 4.2 pp, high confidence |
| Muse Wonder | 50.4% | 78.6% | estimated ± 4.2 pp, high confidence |
| Claude Haiku 5.5 | 48.4% | 75.7% | estimated ± 4.2 pp, high confidence |
| Grok 4.7 | 46.3% | 72.8% | estimated ± 4.2 pp, high confidence |
| Muse Spark 1.3 | 41.6% | 67.2% | estimated ± 4.2 pp, high confidence |
| Grok 4.6 | 41.4% | 67.0% | estimated ± 4.2 pp, high confidence |
| Composer 2.5 | 27.7% | 59.4% | estimated ± 4.2 pp, medium confidence |
Fitted on
| Model | CursorBench 4.0 | SWE-bench Pro |
|---|---|---|
| Claude Fable 5.1 | 51.8% | 81.2% |
| Claude Fable 5.2 | 60.0% | 93.0% |
| Claude Opus 5 | 46.6% | 79.2% |
| Claude Opus 5.5 | 57.8% | 89.9% |
| Claude Sonnet 5 | 34.1% | 63.2% |
| Claude Sonnet 5.5 | 55.5% | 81.3% |
| GPT-5.6 Luna | 35.9% | 62.7% |
| GPT-5.6 Sol | 41.7% | 64.6% |
| GPT-5.6 Terra | 41.3% | 63.4% |