Calibration
FrontierSWE v2 → CursorBench 4.0
CursorBench 4.0 is estimated from FrontierSWE v2 with a Hill curve fitted on 8 models measured on both: y = 0.4198 + (1.2000 − 0.4198)·x^6.00 / (0.79177^6.00 + x^6.00), R² = 0.92, cross-validated error 2.9 pp. It is used for 3 estimates.
| Estimated model | FrontierSWE v2 | CursorBench 4.0 | Source |
|---|---|---|---|
| Gemini 4 Argon | 55.0% | 49.9% | estimated ± 2.9 pp, high confidence |
| GLM-5.3-Flash | 18.1% | 42.0% | estimated ± 2.9 pp, medium confidence |
| Qwen3.8 Max | 15.8% | 42.0% | estimated ± 2.9 pp, medium confidence |