Calibration
Terminal-Bench 2.1 → AA Terminal-Bench 4.0
AA Terminal-Bench 4.0 is estimated from Terminal-Bench 2.1 with a inverse Michaelis–Menten curve fitted on 13 models measured on both: y = 2.01417·(x − 0.6313) / (0.6313 + 2.0000 − x), R² = 0.73, cross-validated error 8.9 pp. It is used for 49 estimates.
| Estimated model | Terminal-Bench 2.1 | AA Terminal-Bench 4.0 | Source |
|---|---|---|---|
| A.X K2 | 36.0% | 0.0% | estimated ± 8.9 pp, low confidence |
| Apodex 1.1 | 70.8% | 8.0% | estimated ± 8.9 pp, medium confidence |
| Atria Dawn Preview | 78.3% | 16.5% | estimated ± 8.9 pp, medium confidence |
| Claude Fable 5 | 84.3% | 23.8% | estimated ± 8.9 pp, medium confidence |
| Claude Mythos 5 | 88.0% | 28.6% | estimated ± 8.9 pp, medium confidence |
| Claude Opus 4.8 | 74.6% | 12.2% | estimated ± 8.9 pp, medium confidence |
| Claude Sonnet 5 | 80.4% | 19.0% | estimated ± 8.9 pp, medium confidence |
| DeepSeek V4 Flash 0731 | 82.7% | 21.8% | estimated ± 8.9 pp, medium confidence |
| DeepSeek V4 Pro 0813 | 87.9% | 28.5% | estimated ± 8.9 pp, medium confidence |
| dots3-note Preview | 75.1% | 12.8% | estimated ± 8.9 pp, medium confidence |
| Ember-1 | 82.0% | 21.0% | estimated ± 8.9 pp, medium confidence |
| Gemini 3.5 Flash | 76.2% | 14.1% | estimated ± 8.9 pp, medium confidence |
| Gemini 3.5 Flash-Lite | 54.0% | 0.0% | estimated ± 8.9 pp, medium confidence |
| Gemini 3.7 Flash | 85.8% | 25.7% | estimated ± 8.9 pp, medium confidence |
| GLM-5.2 | 81.0% | 19.8% | estimated ± 8.9 pp, medium confidence |
| GPT-5.6 Luna | 84.7% | 24.3% | estimated ± 8.9 pp, medium confidence |
| GPT-5.6 Sol | 91.9% | 33.8% | estimated ± 8.9 pp, low confidence |
| GPT-5.6 Terra | 87.4% | 27.8% | estimated ± 8.9 pp, medium confidence |
| Granite 4.2 30B | 29.2% | 0.0% | estimated ± 8.9 pp, low confidence |
| Granite 4.2 8B | 20.6% | 0.0% | estimated ± 8.9 pp, low confidence |
| Grok 4.5 | 83.3% | 22.6% | estimated ± 8.9 pp, medium confidence |
| Hy4 preview | 85.4% | 25.2% | estimated ± 8.9 pp, medium confidence |
| Inkling-Small | 64.7% | 1.6% | estimated ± 8.9 pp, medium confidence |
| K-EXAONE 2.0 | 43.8% | 0.0% | estimated ± 8.9 pp, low confidence |
| Laguna S 2.1 | 70.2% | 7.4% | estimated ± 8.9 pp, medium confidence |
| Ling 3.0 Flash | 57.0% | 0.0% | estimated ± 8.9 pp, medium confidence |
| MAI-Code-1.1-Flash | 62.9% | 0.0% | estimated ± 8.9 pp, medium confidence |
| MiniCPM5-2B | 8.6% | 0.0% | estimated ± 8.9 pp, low confidence |
| Muse Spark 1.1 | 80.0% | 18.5% | estimated ± 8.9 pp, medium confidence |
| Muse Spark 1.2 | 82.9% | 22.1% | estimated ± 8.9 pp, medium confidence |
| Nemotron 3.5 Lightning 30B A3B NVFP4 | 23.5% | 0.0% | estimated ± 8.9 pp, low confidence |
| Ornith-1.0-35B | 64.2% | 1.1% | estimated ± 8.9 pp, medium confidence |
| Ornith-1.0-397B | 77.5% | 15.6% | estimated ± 8.9 pp, medium confidence |
| Ornith-1.0-9B | 43.1% | 0.0% | estimated ± 8.9 pp, low confidence |
| Ornith-1.5-35B-A3B | 67.8% | 4.8% | estimated ± 8.9 pp, medium confidence |
| Ornith-1.5-397B | 86.1% | 26.1% | estimated ± 8.9 pp, medium confidence |
| Ornith-1.5-9B | 46.2% | 0.0% | estimated ± 8.9 pp, low confidence |
| Pokee-Isaac 28B | 65.1% | 2.0% | estimated ± 8.9 pp, medium confidence |
| Quasar 438B | 69.3% | 6.4% | estimated ± 8.9 pp, medium confidence |
| Qwen3.8 Max | 86.6% | 26.8% | estimated ± 8.9 pp, medium confidence |
| Beam | 80.1% | 18.7% | estimated ± 8.9 pp, medium confidence |
| Sakana Fugu | 80.2% | 18.8% | estimated ± 8.9 pp, medium confidence |
| Sakana Fugu-Ultra | 82.1% | 21.1% | estimated ± 8.9 pp, medium confidence |
| Seed 2.1 Pro | 71.0% | 8.2% | estimated ± 8.9 pp, medium confidence |
| Seed 2.1 Turbo | 67.6% | 4.6% | estimated ± 8.9 pp, medium confidence |
| Solar Pro 4 | 57.0% | 0.0% | estimated ± 8.9 pp, medium confidence |
| Step 3.7 Flash | 59.5% | 0.0% | estimated ± 8.9 pp, medium confidence |
| SWE-1.7 | 81.5% | 20.4% | estimated ± 8.9 pp, medium confidence |
| Ternary Bonsai 2 27B | 52.8% | 0.0% | estimated ± 8.9 pp, medium confidence |