Calibration
ITBench SRE → Harvey LAB
Harvey LAB is estimated from ITBench SRE with a Hill curve fitted on 5 models measured on both: y = 0.0000 + (0.9416 − 0.0000)·x^6.00 / (0.20417^6.00 + x^6.00), R² = 0.44, cross-validated error 1.6 pp. It is used for 11 estimates.
| Estimated model | ITBench SRE | Harvey LAB | Source |
|---|---|---|---|
| GPT-6 Astra (max) | 48.6% | 93.6% | estimated ± 1.6 pp, low confidence |
| Gemini 3.8 Flash (high) | 52.5% | 93.8% | estimated ± 1.6 pp, low confidence |
| Muse Spark 1.3 (max) | 33.2% | 89.3% | estimated ± 1.6 pp, low confidence |
| GPT-6 Sol (max) | 49.4% | 93.7% | estimated ± 1.6 pp, low confidence |
| DeepSeek V4.1 Flash (max) | 46.9% | 93.5% | estimated ± 1.6 pp, low confidence |
| Grok 4.7 (xhigh) | 42.1% | 92.9% | estimated ± 1.6 pp, low confidence |
| GPT-5.6 Sol (max) | 56.2% | 93.9% | estimated ± 1.6 pp, low confidence |
| GPT-5.5 (xhigh) | 45.8% | 93.4% | estimated ± 1.6 pp, low confidence |
| GPT-5.6 Terra (max) | 51.0% | 93.8% | estimated ± 1.6 pp, low confidence |
| GLM-5.2 (max) | 42.7% | 93.0% | estimated ± 1.6 pp, low confidence |
| Claude Opus 4.7 (max) | 46.7% | 93.5% | estimated ± 1.6 pp, low confidence |