Calibration
HMMT Feb 2026 → IMOAnswerBench
IMOAnswerBench is estimated from HMMT Feb 2026 with a Michaelis–Menten + offset curve fitted on 8 models measured on both: y = 0.0000 + 2.0000·x / (1.25219 + x), R² = 0.87, cross-validated error 5.9 pp. It is used for 15 estimates.
| Estimated model | HMMT Feb 2026 | IMOAnswerBench | Source |
|---|---|---|---|
| Claude Opus 4.5 | 85.3% | 81.0% | estimated ± 5.9 pp, medium confidence |
| GLM-5 | 86.4% | 81.7% | estimated ± 5.9 pp, medium confidence |
| GLM-5.1 | 82.6% | 79.5% | estimated ± 5.9 pp, medium confidence |
| GLM-5.2 | 92.5% | 85.0% | estimated ± 5.9 pp, medium confidence |
| Inkling-Small | 90.2% | 83.7% | estimated ± 5.9 pp, medium confidence |
| Kimi K2.6 | 92.7% | 85.1% | estimated ± 5.9 pp, medium confidence |
| Kimi K2.5 | 87.1% | 82.0% | estimated ± 5.9 pp, medium confidence |
| MAI-Thinking-1 | 84.9% | 80.8% | estimated ± 5.9 pp, medium confidence |
| MiniCPM5-1B | 25.8% | 34.1% | estimated ± 5.9 pp, low confidence |
| MiniCPM5-2B | 63.8% | 67.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.5 397B | 87.9% | 82.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.6-27B | 84.3% | 80.5% | estimated ± 5.9 pp, medium confidence |
| Qwen3.6-35B-A3B | 83.6% | 80.1% | estimated ± 5.9 pp, medium confidence |
| Qwen3.6 Plus | 87.8% | 82.4% | estimated ± 5.9 pp, medium confidence |
| Solar Open 2 | 93.9% | 85.7% | estimated ± 5.9 pp, medium confidence |