Calibration
SWE-bench Pro → CursorBench 4.0
CursorBench 4.0 is estimated from SWE-bench Pro with a linear curve fitted on 8 models measured on both: y = 0.7664·x + -0.1050, R² = 0.89, cross-validated error 3.6 pp. It is used for 35 estimates.
| Estimated model | SWE-bench Pro | CursorBench 4.0 | Source |
|---|---|---|---|
| Claude Mythos 5 | 80.3% | 51.0% | estimated ± 3.6 pp, high confidence |
| Sakana Fugu-Ultra | 73.7% | 46.0% | estimated ± 3.6 pp, high confidence |
| Beam 1.0 501B-A23B FP8 | 65.9% | 40.0% | estimated ± 3.6 pp, high confidence |
| Hy4 preview | 65.7% | 39.8% | estimated ± 3.6 pp, high confidence |
| Beam 1.0 501B-A23B NVFP4 | 65.6% | 39.8% | estimated ± 3.6 pp, high confidence |
| Ornith-1.5-397B | 65.1% | 39.4% | estimated ± 3.6 pp, high confidence |
| Qwen3.8-Omni-Flash | 63.3% | 38.0% | estimated ± 3.6 pp, high confidence |
| Ornith-1.0-397B | 62.2% | 37.2% | estimated ± 3.6 pp, medium confidence |
| Sage-A1.5 | 61.2% | 36.4% | estimated ± 3.6 pp, medium confidence |
| dots3-note Preview | 61.0% | 36.2% | estimated ± 3.6 pp, medium confidence |
| Atria Dawn Preview | 59.6% | 35.2% | estimated ± 3.6 pp, medium confidence |
| Ornith-1.5-35B-A3B | 59.6% | 35.2% | estimated ± 3.6 pp, medium confidence |
| Laguna S 2.1 | 59.4% | 35.0% | estimated ± 3.6 pp, medium confidence |
| Sakana Fugu | 59.0% | 34.7% | estimated ± 3.6 pp, medium confidence |
| Qwen 3.6 Max (preview) | 57.3% | 33.4% | estimated ± 3.6 pp, medium confidence |
| Claude Opus 4.5 | 57.1% | 33.3% | estimated ± 3.6 pp, medium confidence |
| GPT-5.3 Codex | 56.8% | 33.0% | estimated ± 3.6 pp, medium confidence |
| Qwen3.6 Plus | 56.6% | 32.9% | estimated ± 3.6 pp, medium confidence |
| MiMo-V2.5 | 56.1% | 32.5% | estimated ± 3.6 pp, medium confidence |
| GPT-5.2 | 55.6% | 32.1% | estimated ± 3.6 pp, medium confidence |
| GLM-5 | 55.1% | 31.7% | estimated ± 3.6 pp, medium confidence |
| Spark-X2.6-4B | 54.3% | 31.1% | estimated ± 3.6 pp, medium confidence |
| Claude Opus 4.6 | 53.4% | 30.4% | estimated ± 3.6 pp, medium confidence |
| MAI-Thinking-1 | 52.8% | 30.0% | estimated ± 3.6 pp, medium confidence |
| Muse Spark | 52.4% | 29.7% | estimated ± 3.6 pp, medium confidence |
| Grok 4.20 | 51.8% | 29.2% | estimated ± 3.6 pp, medium confidence |
| Kimi K2.5 | 50.7% | 28.4% | estimated ± 3.6 pp, medium confidence |
| Ornith-1.0-35B | 50.4% | 28.1% | estimated ± 3.6 pp, medium confidence |
| Qwen3.6-35B-A3B | 49.5% | 27.4% | estimated ± 3.6 pp, medium confidence |
| Laguna M.1 | 49.2% | 27.2% | estimated ± 3.6 pp, medium confidence |
| Laguna XS 2.1 | 47.6% | 26.0% | estimated ± 3.6 pp, medium confidence |
| Ornith-1.5-9B | 47.5% | 25.9% | estimated ± 3.6 pp, medium confidence |
| Laguna XS.2 | 46.3% | 25.0% | estimated ± 3.6 pp, medium confidence |
| Ornith-1.0-9B | 42.9% | 22.4% | estimated ± 3.6 pp, medium confidence |
| LongCat-Flash-Lite-Sparse | 40.6% | 20.6% | estimated ± 3.6 pp, medium confidence |
Fitted on
| Model | SWE-bench Pro | CursorBench 4.0 |
|---|---|---|
| Claude Fable 5.1 | 81.2% | 51.8% |
| Claude Opus 5 | 79.2% | 46.6% |
| Claude Opus 5.5 | 89.9% | 57.8% |
| Claude Sonnet 5 | 63.2% | 34.1% |
| Claude Sonnet 5.5 | 81.3% | 55.5% |
| GPT-5.6 Luna | 62.7% | 35.9% |
| GPT-5.6 Sol | 64.6% | 41.7% |
| GPT-5.6 Terra | 63.4% | 41.3% |