benchgap
Calibration

HMMT Feb 2026 → AIME26

AIME26 is estimated from HMMT Feb 2026 with a Hill curve fitted on 19 models measured on both: y = 0.0000 + (1.0714 − 0.0000)·x^2.10 / (0.32639^2.10 + x^2.10), R² = 0.99, cross-validated error 1.3 pp. It is used for 4 estimates.

Estimated modelHMMT Feb 2026AIME26Source
DeepSeek V4 Flash 073194.8%96.8%estimated ± 1.3 pp, medium confidence
DeepSeek V4 Pro 081395.2%96.9%estimated ± 1.3 pp, medium confidence
Qwen3.7 Max97.1%97.3%estimated ± 1.3 pp, medium confidence
Qwen3.7 Plus92.9%96.4%estimated ± 1.3 pp, high confidence