Math

HMMT Feb 2026 leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on HMMT Feb 2026 is 97.1% by Qwen3.7 Max. 14 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	dots3-note Preview	97.3%	estimated ± 9.7 pp, low confidence
2	Qwen3.7 Max	97.1%	measured
3	DeepSeek V4 Pro 0813	95.2%	measured
4	DeepSeek V4 Flash 0731	94.8%	measured
5	Solar Open 2	93.9%	measured
6	Qwen3.7 Plus	92.9%	measured
7	Kimi K2.6	92.7%	measured
8	GLM-5.2	92.5%	measured
9	Inkling-Small	90.2%	measured
10	Beam	89.8%	estimated ± 4.0 pp, high confidence
11	A.X K2	88.9%	estimated ± 4.0 pp, high confidence
12	Inkling	88.9%	estimated ± 4.0 pp, high confidence
13	Qwen3.5 397B	87.9%	measured
14	Qwen3.6 Plus	87.8%	measured
15	Ternary Bonsai 2 27B	87.2%	estimated ± 4.0 pp, high confidence
16	Kimi K2.5	87.1%	measured
17	Ling 3.0 Flash	87.0%	measured
18	Solar Pro 4	86.5%	estimated ± 4.0 pp, high confidence
19	GLM-5	86.4%	measured
20	Muse Glimmer 30B	85.6%	estimated ± 4.0 pp, high confidence
21	Claude Opus 4.5	85.3%	measured
22	MAI-Thinking-1	84.9%	measured
23	Qwen3.6-27B	84.3%	measured
24	Qwen3.6-35B-A3B	83.6%	measured
25	Granite 4.2 30B	83.2%	estimated ± 1.2 pp, low confidence
26	GLM-5.1	82.6%	measured
27	K-EXAONE 2.0	78.4%	measured
28	Granite 4.2 8B	77.0%	estimated ± 1.2 pp, low confidence
29	ZAYA1-8B	71.6%	measured
30	Granite 4.2 3B	69.5%	estimated ± 1.2 pp, low confidence
31	MiniCPM5-2B	63.8%	measured
32	Gemma 4 12B	57.5%	estimated ± 4.0 pp, high confidence
33	ZAYA1-74B-Preview	55.7%	estimated ± 4.0 pp, high confidence
34	LongCat-Flash-Lite-Sparse	40.5%	measured
35	LFM2.5-8B-A1B	27.3%	estimated ± 4.0 pp, high confidence
36	MiniCPM5-1B	25.8%	measured
37	LLaDA2.2-mini	24.2%	estimated ± 4.0 pp, medium confidence

HMMT Feb 2026 leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General