benchgap
Calibration

AIME26 → HMMT Feb 2026

HMMT Feb 2026 is estimated from AIME26 with a Hill curve fitted on 19 models measured on both: y = 0.2374 + (1.2000 − 0.2374)·x^6.00 / (0.85857^6.00 + x^6.00), R² = 0.95, cross-validated error 4.0 pp. It is used for 10 estimates.

Estimated modelAIME26HMMT Feb 2026Source
A.X K297.1%88.9%estimated ± 4.0 pp, high confidence
Gemma 4 12B77.5%57.5%estimated ± 4.0 pp, high confidence
Inkling97.1%88.9%estimated ± 4.0 pp, high confidence
LFM2.5-8B-A1B50.0%27.3%estimated ± 4.0 pp, high confidence
LLaDA2.2-mini35.1%24.2%estimated ± 4.0 pp, medium confidence
Muse Glimmer 30B94.7%85.6%estimated ± 4.0 pp, high confidence
Beam97.8%89.8%estimated ± 4.0 pp, high confidence
Solar Pro 495.3%86.5%estimated ± 4.0 pp, high confidence
Ternary Bonsai 2 27B95.8%87.2%estimated ± 4.0 pp, high confidence
ZAYA1-74B-Preview76.4%55.7%estimated ± 4.0 pp, high confidence