Calibration
AA-HLE → LABBench2
LABBench2 is estimated from AA-HLE with a linear curve fitted on 7 models measured on both: y = 0.4257·x + 0.6276, R² = 0.64, cross-validated error 2.4 pp. It is used for 25 estimates.
| Estimated model | AA-HLE | LABBench2 | Source |
|---|---|---|---|
| Claude Sonnet 5.5 | 55.0% | 86.2% | estimated ± 2.4 pp, medium confidence |
| Muse Wonder Full | 53.6% | 85.6% | estimated ± 2.4 pp, medium confidence |
| Kimi K3.1 Pro | 52.5% | 85.1% | estimated ± 2.4 pp, medium confidence |
| Muse Wonder | 50.1% | 84.1% | estimated ± 2.4 pp, medium confidence |
| MiMo-V2.6-Pro | 49.4% | 83.8% | estimated ± 2.4 pp, medium confidence |
| MiniMax M3.1 Flash | 48.0% | 83.2% | estimated ± 2.4 pp, medium confidence |
| Step 5 Preview | 46.5% | 82.6% | estimated ± 2.4 pp, medium confidence |
| GPT-6.1 Terra | 46.2% | 82.4% | estimated ± 2.4 pp, medium confidence |
| Claude Haiku 5.5 | 44.4% | 81.7% | estimated ± 2.4 pp, medium confidence |
| Grok 4.7 | 43.1% | 81.1% | estimated ± 2.4 pp, medium confidence |
| Ling 3.1 Flash | 39.4% | 79.5% | estimated ± 2.4 pp, low confidence |
| DeepSeek V4.1 Flash 1017 | 39.2% | 79.5% | estimated ± 2.4 pp, low confidence |
| GPT-6 Luna | 38.5% | 79.2% | estimated ± 2.4 pp, low confidence |
| MiMo-V2.6-Flash | 35.1% | 77.7% | estimated ± 2.4 pp, low confidence |
| Beam 1.0 501B-A23B | 34.7% | 77.5% | estimated ± 2.4 pp, low confidence |
| Apodex 1.1 | 34.1% | 77.3% | estimated ± 2.4 pp, low confidence |
| Apodex 1.1 Mini | 34.1% | 77.3% | estimated ± 2.4 pp, low confidence |
| Nemotron 3.5 Super VL 120B A12B | 30.7% | 75.8% | estimated ± 2.4 pp, low confidence |
| Nemotron 3.5 Super VL 120B A12B NVFP4 | 30.7% | 75.8% | estimated ± 2.4 pp, low confidence |
| A.X K2 | 29.6% | 75.4% | estimated ± 2.4 pp, low confidence |
| GLM-5 | 29.3% | 75.2% | estimated ± 2.4 pp, low confidence |
| GPT-5.4 nano | 28.3% | 74.8% | estimated ± 2.4 pp, low confidence |
| GLM-5-Turbo | 27.8% | 74.6% | estimated ± 2.4 pp, low confidence |
| GPT-5.1-Codex | 25.7% | 73.7% | estimated ± 2.4 pp, low confidence |
| GPT-5.1-Codex-Max | 25.7% | 73.7% | estimated ± 2.4 pp, low confidence |
Fitted on
| Model | AA-HLE | LABBench2 |
|---|---|---|
| Claude Opus 5 | 54.9% | 84.2% |
| Claude Sonnet 5 | 41.3% | 80.1% |
| Gemini 3.7 Flash | 47.9% | 82.1% |
| Gemini 3.8 Flash | 47.8% | 86.2% |
| Gemini 4 Argon | 57.1% | 88.8% |
| GPT-5.6 Sol | 49.5% | 82.1% |
| GPT-5.6 Terra | 42.9% | 81.2% |