Calibration
Terminal-Bench 3.0 → Terminal-Bench 2.1
Terminal-Bench 2.1 is estimated from Terminal-Bench 3.0 with a linear curve fitted on 11 models measured on both: y = 0.3715·x + 0.7596, R² = 0.31, cross-validated error 6.6 pp. It is used for 0 estimates.
Fitted on
| Model | Terminal-Bench 3.0 | Terminal-Bench 2.1 |
|---|---|---|
| Claude Fable 5 | 34.1% | 84.3% |
| Claude Opus 4.8 | 21.1% | 74.6% |
| Claude Sonnet 5 | 14.6% | 80.4% |
| Gemini 3.7 Flash | 14.9% | 85.8% |
| GLM-5.2 | 4.6% | 81.0% |
| GPT-5.6 Luna | 14.3% | 84.7% |
| GPT-5.6 Sol | 34.6% | 91.9% |
| GPT-5.6 Terra | 20.8% | 87.4% |
| Grok 4.5 | 15.7% | 83.3% |
| Ornith-1.5-35B-A3B | 5.1% | 67.8% |
| Ornith-1.5-397B | 13.5% | 86.1% |