Calibration
FrontierMath v2 (Tiers 1-3) → HMMT Nov 2025
HMMT Nov 2025 is estimated from FrontierMath v2 (Tiers 1-3) with a Michaelis–Menten curve fitted on 6 models measured on both: y = 0.9827·x / (0.02058 + x), R² = 0.69, cross-validated error 1.5 pp. It is used for 63 estimates.
| Estimated model | FrontierMath v2 (Tiers 1-3) | HMMT Nov 2025 | Source |
|---|---|---|---|
| Claude Fable 5.2 | 95.1% | 96.2% | estimated ± 1.5 pp, low confidence |
| GPT-6.1 Sol | 93.7% | 96.2% | estimated ± 1.5 pp, low confidence |
| GPT-6 Astra | 93.7% | 96.2% | estimated ± 1.5 pp, low confidence |
| GPT-6.2 Astra | 92.6% | 96.1% | estimated ± 1.5 pp, low confidence |
| Claude Opus 5.5 | 91.2% | 96.1% | estimated ± 1.5 pp, low confidence |
| GPT-6.1 Terra | 90.9% | 96.1% | estimated ± 1.5 pp, low confidence |
| Gemini 4 Carbon | 90.5% | 96.1% | estimated ± 1.5 pp, low confidence |
| Claude Fable 5.1 | 90.2% | 96.1% | estimated ± 1.5 pp, low confidence |
| GPT-6 Sol | 89.8% | 96.1% | estimated ± 1.5 pp, low confidence |
| GPT-5.6 Sol | 89.0% | 96.1% | estimated ± 1.5 pp, low confidence |
| Claude Sonnet 5.5 | 88.8% | 96.0% | estimated ± 1.5 pp, low confidence |
| GPT-5.5 Pro | 87.7% | 96.0% | estimated ± 1.5 pp, low confidence |
| Claude Fable 5 | 87.0% | 96.0% | estimated ± 1.5 pp, low confidence |
| Gemini 4 Argon | 86.0% | 96.0% | estimated ± 1.5 pp, low confidence |
| Claude Opus 5 | 85.6% | 96.0% | estimated ± 1.5 pp, low confidence |
| GPT-5.5 | 85.3% | 96.0% | estimated ± 1.5 pp, low confidence |
| GPT-5.6 Terra | 84.9% | 95.9% | estimated ± 1.5 pp, low confidence |
| GPT-5.4 Pro | 82.5% | 95.9% | estimated ± 1.5 pp, low confidence |
| Claude Opus 4.8 | 80.0% | 95.8% | estimated ± 1.5 pp, low confidence |
| GPT-6 Luna | 78.9% | 95.8% | estimated ± 1.5 pp, low confidence |
| GPT-5.6 Luna | 78.6% | 95.8% | estimated ± 1.5 pp, low confidence |
| GPT-5.4 | 78.6% | 95.8% | estimated ± 1.5 pp, low confidence |
| Claude Haiku 5.5 | 75.1% | 95.7% | estimated ± 1.5 pp, low confidence |
| Qwen3.8 Max | 74.7% | 95.6% | estimated ± 1.5 pp, low confidence |
| Muse Spark 1.3 | 74.0% | 95.6% | estimated ± 1.5 pp, low confidence |
| Gemini 3.7 Flash | 71.6% | 95.5% | estimated ± 1.5 pp, low confidence |
| Claude Opus 4.7 | 70.2% | 95.5% | estimated ± 1.5 pp, low confidence |
| GLM-5.3 | 68.8% | 95.4% | estimated ± 1.5 pp, low confidence |
| Gemini 3.8 Flash | 68.4% | 95.4% | estimated ± 1.5 pp, low confidence |
| GPT-5.2 | 67.4% | 95.4% | estimated ± 1.5 pp, low confidence |
| DeepSeek V4.1 Flash 1017 | 67.4% | 95.4% | estimated ± 1.5 pp, low confidence |
| Claude Opus 4.6 | 66.0% | 95.3% | estimated ± 1.5 pp, low confidence |
| Grok 4.6 | 66.0% | 95.3% | estimated ± 1.5 pp, low confidence |
| Claude Sonnet 5 | 65.6% | 95.3% | estimated ± 1.5 pp, low confidence |
| DeepSeek V4 Pro | 64.6% | 95.2% | estimated ± 1.5 pp, low confidence |
| Qwen3.7 Max | 64.6% | 95.2% | estimated ± 1.5 pp, low confidence |
| Gemini 3.5 Flash | 62.8% | 95.2% | estimated ± 1.5 pp, low confidence |
| Gemini 3.1 Pro | 59.6% | 95.0% | estimated ± 1.5 pp, low confidence |
| Gemini 3.6 Flash | 58.9% | 95.0% | estimated ± 1.5 pp, medium confidence |
| Grok 4.5 | 57.2% | 94.9% | estimated ± 1.5 pp, medium confidence |
| Kimi K2.6 | 57.2% | 94.9% | estimated ± 1.5 pp, medium confidence |
| GLM-5.3-Flash | 55.8% | 94.8% | estimated ± 1.5 pp, medium confidence |
| Kimi K2.7 Code | 54.0% | 94.7% | estimated ± 1.5 pp, medium confidence |
| Grok 4.7 | 53.0% | 94.6% | estimated ± 1.5 pp, medium confidence |
| Gemini 3 Flash | 51.2% | 94.5% | estimated ± 1.5 pp, medium confidence |
| GPT-5.4 mini | 51.2% | 94.5% | estimated ± 1.5 pp, medium confidence |
| Inkling-Small | 46.3% | 94.1% | estimated ± 1.5 pp, medium confidence |
| GPT-5.4 nano | 44.9% | 94.0% | estimated ± 1.5 pp, medium confidence |
| Grok 4.3 | 42.8% | 93.8% | estimated ± 1.5 pp, medium confidence |
| o4-mini | 36.1% | 93.0% | estimated ± 1.5 pp, medium confidence |
| Qwen3.7 Plus | 34.4% | 92.7% | estimated ± 1.5 pp, medium confidence |
| Inkling | 33.3% | 92.6% | estimated ± 1.5 pp, medium confidence |
| o3 | 33.3% | 92.6% | estimated ± 1.5 pp, medium confidence |
| Gemini 3.1 Flash-Lite | 27.7% | 91.5% | estimated ± 1.5 pp, medium confidence |
| Gemini 3.5 Flash-Lite | 26.0% | 91.1% | estimated ± 1.5 pp, medium confidence |
| Gemini 2.5 Pro | 24.6% | 90.7% | estimated ± 1.5 pp, medium confidence |
| Claude Sonnet 4.5 | 23.9% | 90.5% | estimated ± 1.5 pp, medium confidence |
| Qwen3.7 Flash | 19.3% | 88.8% | estimated ± 1.5 pp, low confidence |
| Qwen3.5 Flash | 18.2% | 88.3% | estimated ± 1.5 pp, low confidence |
| o1 | 14.7% | 86.2% | estimated ± 1.5 pp, low confidence |
| GPT-4.1 mini | 6.7% | 75.1% | estimated ± 1.5 pp, low confidence |
| GPT-4.1 | 6.0% | 73.1% | estimated ± 1.5 pp, low confidence |
| GPT-4o | 0.4% | 14.3% | estimated ± 1.5 pp, low confidence |
Fitted on
| Model | FrontierMath v2 (Tiers 1-3) | HMMT Nov 2025 |
|---|---|---|
| Claude Opus 4.5 | 34.4% | 93.3% |
| GLM-5.1 | 36.8% | 94.0% |
| GLM-5.2 | 59.2% | 94.4% |
| Qwen3.6-27B | 35.1% | 90.7% |
| Qwen3.6-35B-A3B | 20.4% | 89.1% |
| Qwen3.6 Plus | 38.2% | 94.6% |