Calibration
Terminal-Bench-Science 0.1 → Terminal-Bench 4.0
Terminal-Bench 4.0 is estimated from Terminal-Bench-Science 0.1 with a Michaelis–Menten curve fitted on 11 models measured on both: y = 0.7823·x / (0.14226 + x), R² = 0.89, cross-validated error 6.9 pp. It is used for 0 estimates.
Fitted on
| Model | Terminal-Bench-Science 0.1 | Terminal-Bench 4.0 |
|---|---|---|
| Claude Fable 5.1 | 52.6% | 55.8% |
| Claude Fable 5.2 | 74.6% | 67.6% |
| Claude Mythos 5.2 | 75.3% | 70.5% |
| Claude Opus 5.5 | 58.7% | 66.4% |
| Claude Sonnet 5.5 | 59.9% | 70.6% |
| ERNIE 5.5 Flash Preview | 21.0% | 46.5% |
| GPT-6.1 Terra | 58.3% | 50.3% |
| GPT-6 Astra | 64.6% | 57.9% |
| Hy4 | 19.8% | 44.4% |
| Inkling 2 Mini | 1.4% | 5.6% |
| Kimi K3.1 Pro | 38.1% | 65.7% |