benchgap
Math

HMMT Nov 2025 leaderboard

As of 2026-11-21, the highest measured score on HMMT Nov 2025 is 96.9% by GLM-5. 88 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GLM-596.9%measured
2Inkling 2 Small96.5%estimated ± 2.8 pp, medium confidence
3DeepSeek V4 Flash 073196.2%estimated ± 2.8 pp, medium confidence
4Claude Fable 5.296.2%estimated ± 1.5 pp, low confidence
5Beam 1.0 501B-A23B FP896.2%estimated ± 3.2 pp, high confidence
6GPT-6.1 Sol96.2%estimated ± 1.5 pp, low confidence
7GPT-6 Astra96.2%estimated ± 1.5 pp, low confidence
8GPT-6.2 Astra96.1%estimated ± 1.5 pp, low confidence
9Claude Opus 5.596.1%estimated ± 1.5 pp, low confidence
10Gemini 4 Neon96.1%estimated ± 1.5 pp, low confidence
11GPT-6.1 Terra96.1%estimated ± 1.5 pp, low confidence
12Gemini 4 Carbon96.1%estimated ± 1.5 pp, low confidence
13Claude Fable 5.196.1%estimated ± 1.5 pp, low confidence
14GPT-6 Sol96.1%estimated ± 1.5 pp, low confidence
15GPT-5.6 Sol96.1%estimated ± 1.5 pp, low confidence
16Claude Sonnet 5.596.0%estimated ± 1.5 pp, low confidence
17Gemini 4 Hydrogen96.0%estimated ± 1.5 pp, low confidence
18GPT-5.5 Pro96.0%estimated ± 1.5 pp, low confidence
19Claude Fable 596.0%estimated ± 1.5 pp, low confidence
20Gemini 4 Argon96.0%estimated ± 1.5 pp, low confidence
21Claude Opus 596.0%estimated ± 1.5 pp, low confidence
22Solar Open 296.0%estimated ± 2.8 pp, medium confidence
23GPT-5.596.0%estimated ± 1.5 pp, low confidence
24GPT-5.6 Terra95.9%estimated ± 1.5 pp, low confidence
25GPT-5.4 Pro95.9%estimated ± 1.5 pp, low confidence
26Beam 1.0 501B-A23B95.8%estimated ± 3.2 pp, high confidence
27Claude Opus 4.895.8%estimated ± 1.5 pp, low confidence
28GPT-6 Luna95.8%estimated ± 1.5 pp, low confidence
29GPT-5.6 Luna95.8%estimated ± 1.5 pp, low confidence
30GPT-5.495.8%estimated ± 1.5 pp, low confidence
31Claude Haiku 5.595.7%estimated ± 1.5 pp, low confidence
32Qwen3.8 Max95.6%estimated ± 1.5 pp, low confidence
33Muse Spark 1.395.6%estimated ± 1.5 pp, low confidence
34Beam 1.0 501B-A23B NVFP495.6%estimated ± 3.2 pp, high confidence
35Gemini 3.7 Flash95.5%estimated ± 1.5 pp, low confidence
36Claude Opus 4.795.5%estimated ± 1.5 pp, low confidence
37GLM-5.395.4%estimated ± 1.5 pp, low confidence
38Gemini 3.8 Flash95.4%estimated ± 1.5 pp, low confidence
39GPT-5.295.4%estimated ± 1.5 pp, low confidence
40DeepSeek V4.1 Flash 101795.4%estimated ± 1.5 pp, low confidence
41Claude Opus 4.695.3%estimated ± 1.5 pp, low confidence
42Grok 4.695.3%estimated ± 1.5 pp, low confidence
43Claude Sonnet 595.3%estimated ± 1.5 pp, low confidence
44DeepSeek V4 Pro95.2%estimated ± 1.5 pp, low confidence
45Qwen3.7 Max95.2%estimated ± 1.5 pp, low confidence
46Gemini 3.5 Flash95.2%estimated ± 1.5 pp, low confidence
47A.X K295.1%estimated ± 3.2 pp, high confidence
48Gemini 3.1 Pro95.0%estimated ± 1.5 pp, low confidence
49Gemini 3.6 Flash95.0%estimated ± 1.5 pp, medium confidence
50GLM-4.794.9%estimated ± 1.5 pp, high confidence
51Grok 4.594.9%estimated ± 1.5 pp, medium confidence
52Kimi K2.694.9%estimated ± 1.5 pp, medium confidence
53Ripple-65B Preview94.8%estimated ± 2.8 pp, high confidence
54GLM-5.3-Flash94.8%estimated ± 1.5 pp, medium confidence
55Kimi K2.7 Code94.7%estimated ± 1.5 pp, medium confidence
56Qwen3.6 Plus94.6%measured
57Grok 4.794.6%estimated ± 1.5 pp, medium confidence
58Gemini 3 Flash94.5%estimated ± 1.5 pp, medium confidence
59GPT-5.4 mini94.5%estimated ± 1.5 pp, medium confidence
60GLM-5.294.4%measured
61Inkling-Small94.1%estimated ± 1.5 pp, medium confidence
62GLM-5.194.0%measured
63GPT-5.4 nano94.0%estimated ± 1.5 pp, medium confidence
64Grok 4.393.8%estimated ± 1.5 pp, medium confidence
65Ternary Bonsai 2 27B93.5%estimated ± 3.2 pp, high confidence
66Ling 3.0 Flash93.3%estimated ± 2.8 pp, high confidence
67Claude Opus 4.593.3%measured
68o4-mini93.0%estimated ± 1.5 pp, medium confidence
69Solar Pro 492.8%estimated ± 3.2 pp, high confidence
70Qwen3.7 Plus92.7%estimated ± 1.5 pp, medium confidence
71Qwen3.5 397B92.7%measured
72Inkling92.6%estimated ± 1.5 pp, medium confidence
73o392.6%estimated ± 1.5 pp, medium confidence
74MAI-Thinking-192.3%estimated ± 2.8 pp, high confidence
75Muse Glimmer 30B92.0%estimated ± 3.2 pp, high confidence
76Nemotron 3 Super 100B91.9%estimated ± 1.5 pp, high confidence
77Gemini 3.1 Flash-Lite91.5%estimated ± 1.5 pp, medium confidence
78Kimi K2.591.1%measured
79Gemini 3.5 Flash-Lite91.1%estimated ± 1.5 pp, medium confidence
80Qwen3.6-27B90.7%measured
81Gemini 2.5 Pro90.7%estimated ± 1.5 pp, medium confidence
82Claude Sonnet 4.590.5%estimated ± 1.5 pp, medium confidence
83Spark-X2.6-4B90.0%estimated ± 2.8 pp, high confidence
84Qwen3.6-35B-A3B89.1%measured
85Qwen3.7 Flash88.8%estimated ± 1.5 pp, low confidence
86Qwen3.5 Flash88.3%estimated ± 1.5 pp, low confidence
87Granite 4.2 30B88.1%estimated ± 1.5 pp, high confidence
88K-EXAONE 2.087.7%estimated ± 2.8 pp, high confidence
89o186.2%estimated ± 1.5 pp, low confidence
90Sarvam 105B85.8%measured
91MiniMax M2.781.0%measured
92ZAYA1-8B80.1%estimated ± 2.8 pp, medium confidence
93Granite 4.2 8B78.7%estimated ± 1.5 pp, high confidence
94GPT-4.1 mini75.1%estimated ± 1.5 pp, low confidence
95Sarvam 30B74.2%measured
96GPT-4.173.1%estimated ± 1.5 pp, low confidence
97Granite 4.2 3B68.7%estimated ± 1.5 pp, medium confidence
98MiniCPM5-2B66.5%estimated ± 2.8 pp, medium confidence
99DeepSeek-R1-0528-Qwen3-8B64.2%estimated ± 1.5 pp, medium confidence
100GPT-4o14.3%estimated ± 1.5 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General