benchgap
Math

HMMT Feb 2026 leaderboard

As of 2026-11-21, the highest measured score on HMMT Feb 2026 is 97.1% by Qwen3.7 Max. 112 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5100.0%estimated ± 3.1 pp, medium confidence
2Claude Fable 5.1100.0%estimated ± 3.1 pp, medium confidence
3Claude Fable 5.2100.0%estimated ± 3.1 pp, medium confidence
4Claude Haiku 5.5100.0%estimated ± 1.5 pp, low confidence
5Claude Opus 4.8100.0%estimated ± 1.5 pp, low confidence
6Claude Opus 5100.0%estimated ± 3.1 pp, medium confidence
7Claude Opus 5.5100.0%estimated ± 3.1 pp, medium confidence
8Claude Sonnet 5.5100.0%estimated ± 3.1 pp, medium confidence
9Gemini 4 Argon100.0%estimated ± 3.1 pp, medium confidence
10Gemini 4 Carbon100.0%estimated ± 3.1 pp, medium confidence
11Gemini 4 Hydrogen100.0%estimated ± 3.1 pp, medium confidence
12Gemini 4 Neon100.0%estimated ± 3.1 pp, medium confidence
13GPT-5.4100.0%estimated ± 1.5 pp, low confidence
14GPT-5.4 Pro100.0%estimated ± 1.5 pp, low confidence
15GPT-5.5100.0%estimated ± 3.1 pp, medium confidence
16GPT-5.5 Pro100.0%estimated ± 3.1 pp, medium confidence
17GPT-5.6 Luna100.0%estimated ± 1.5 pp, low confidence
18GPT-5.6 Sol100.0%estimated ± 3.1 pp, medium confidence
19GPT-5.6 Terra100.0%estimated ± 3.1 pp, medium confidence
20GPT-6.1 Sol100.0%estimated ± 3.1 pp, medium confidence
21GPT-6.1 Terra100.0%estimated ± 3.1 pp, medium confidence
22GPT-6.2 Astra100.0%estimated ± 3.1 pp, medium confidence
23GPT-6 Astra100.0%estimated ± 3.1 pp, medium confidence
24GPT-6 Luna100.0%estimated ± 1.5 pp, low confidence
25GPT-6 Sol100.0%estimated ± 3.1 pp, medium confidence
26Muse Spark 1.3100.0%estimated ± 1.5 pp, low confidence
27Qwen3.8 Max100.0%estimated ± 1.5 pp, low confidence
28dots3-note Preview97.5%estimated ± 8.9 pp, low confidence
29Gemini 3.7 Flash97.3%estimated ± 1.5 pp, low confidence
30Qwen3.7 Max97.1%measured
31Inkling 2 Small95.8%measured
32Claude Opus 4.795.4%estimated ± 1.5 pp, medium confidence
33Grok 4.695.4%estimated ± 1.5 pp, medium confidence
34GPT-5.295.4%estimated ± 1.5 pp, medium confidence
35DeepSeek V4 Pro95.2%measured
36DeepSeek V4 Flash 073194.8%measured
37Claude Sonnet 594.5%estimated ± 1.5 pp, medium confidence
38GLM-5.394.5%estimated ± 1.5 pp, medium confidence
39Solar Open 293.9%measured
40Beam 1.0 501B-A23B FP893.7%estimated ± 4.0 pp, high confidence
41Claude Opus 4.693.6%estimated ± 1.5 pp, medium confidence
42DeepSeek V4.1 Flash 101793.6%estimated ± 1.5 pp, medium confidence
43Gemini 3.1 Pro93.6%estimated ± 1.5 pp, medium confidence
44Gemini 3.5 Flash93.6%estimated ± 1.5 pp, medium confidence
45Beam 1.0 501B-A23B93.0%estimated ± 4.0 pp, high confidence
46Qwen3.7 Plus92.9%measured
47Kimi K2.692.7%measured
48Grok 4.592.7%estimated ± 1.5 pp, medium confidence
49Beam 1.0 501B-A23B NVFP492.5%estimated ± 4.0 pp, high confidence
50GLM-5.292.5%measured
51Gemini 3.6 Flash91.8%estimated ± 1.5 pp, medium confidence
52Gemini 3.8 Flash91.8%estimated ± 1.5 pp, medium confidence
53A.X K291.6%estimated ± 4.0 pp, high confidence
54Ripple-65B Preview90.5%measured
55Inkling-Small90.2%measured
56Gemini 3 Flash89.9%estimated ± 1.5 pp, medium confidence
57GLM-5.3-Flash89.9%estimated ± 1.5 pp, medium confidence
58Grok 4.789.9%estimated ± 1.5 pp, medium confidence
59Grok 4.389.0%estimated ± 1.5 pp, medium confidence
60Ternary Bonsai 2 27B88.6%estimated ± 4.0 pp, high confidence
61GPT-5.4 nano88.1%estimated ± 1.5 pp, medium confidence
62Kimi K2.7 Code88.1%estimated ± 1.5 pp, medium confidence
63Qwen3.5 397B87.9%measured
64Step 3.5 Flash87.8%estimated ± 11.0 pp, low confidence
65Qwen3.6 Plus87.8%measured
66Solar Pro 487.4%estimated ± 4.0 pp, high confidence
67GLM-4.787.4%estimated ± 1.2 pp, medium confidence
68GPT-5.4 mini87.2%estimated ± 1.5 pp, medium confidence
69Kimi K2.587.1%measured
70Ling 3.0 Flash87.0%measured
71GLM-586.4%measured
72o386.2%estimated ± 3.1 pp, high confidence
73Muse Glimmer 30B86.1%estimated ± 4.0 pp, high confidence
74OrcaCyber-Nano-1.085.8%estimated ± 11.0 pp, low confidence
75Nemotron 3 Super 100B85.6%estimated ± 1.2 pp, medium confidence
76Inkling85.4%estimated ± 1.5 pp, medium confidence
77o4-mini85.4%estimated ± 1.5 pp, medium confidence
78Claude Opus 4.585.3%measured
79MAI-Thinking-184.9%measured
80Gemini 3.1 Flash-Lite84.6%estimated ± 3.1 pp, high confidence
81Claude Sonnet 4.584.5%estimated ± 1.5 pp, low confidence
82Qwen3.6-27B84.3%measured
83Qwen3.6-35B-A3B83.6%measured
84Gemini 2.5 Pro83.5%estimated ± 1.5 pp, low confidence
85Gemini 3.5 Flash-Lite83.5%estimated ± 1.5 pp, low confidence
86Granite 4.2 30B83.2%estimated ± 1.2 pp, low confidence
87GLM-5.182.6%measured
88Qwen3.7 Flash82.2%estimated ± 3.1 pp, medium confidence
89Qwen3.5 Flash81.9%estimated ± 3.1 pp, medium confidence
90MiMo-V2-Flash81.5%estimated ± 11.0 pp, low confidence
91Sarvam 105B81.3%estimated ± 1.2 pp, low confidence
92Spark-X2.6-4B81.3%measured
93o180.9%estimated ± 3.1 pp, medium confidence
94GPT-4.1 mini78.6%estimated ± 3.1 pp, medium confidence
95GPT-4.178.4%estimated ± 3.1 pp, medium confidence
96K-EXAONE 2.078.4%measured
97GPT-OSS 120B78.0%estimated ± 11.0 pp, low confidence
98GPT-4o76.8%estimated ± 3.1 pp, medium confidence
99GPT-OSS 20B76.2%estimated ± 11.0 pp, low confidence
100GLM-4.7-Flash75.9%estimated ± 11.0 pp, low confidence
101MiniMax M2.772.7%measured
102ZAYA1-8B71.6%measured
103Nemotron 3 Nano 30B70.1%estimated ± 11.0 pp, low confidence
104DeepSeek-R1-052866.2%estimated ± 11.0 pp, low confidence
105Granite 4.2 8B64.2%estimated ± 11.0 pp, low confidence
106MiniCPM5-2B63.8%measured
107Sarvam 30B62.3%estimated ± 3.4 pp, medium confidence
108Exaone 4.0 32B61.0%estimated ± 11.0 pp, low confidence
109Gemma 4 12B55.9%estimated ± 4.0 pp, high confidence
110ZAYA1-74B-Preview54.4%estimated ± 4.0 pp, high confidence
111Nemotron 3 Nano Omni 30B A3B53.8%estimated ± 11.0 pp, low confidence
112Qwen3 235B-A22B (Reasoning)52.5%estimated ± 11.0 pp, low confidence
113Granite 4.2 3B46.3%estimated ± 11.0 pp, low confidence
114Phi-4-reasoning-plus45.7%estimated ± 11.0 pp, low confidence
115MiniMax M1 80k43.8%estimated ± 11.0 pp, low confidence
116DeepSeek-R1-0528-Qwen3-8B42.9%estimated ± 11.0 pp, low confidence
117Claude 4 Opus Thinking41.6%estimated ± 11.0 pp, low confidence
118LongCat-Flash-Lite-Sparse40.5%measured
119Qwen3 32B (Reasoning)38.1%estimated ± 11.0 pp, low confidence
120Nemotron Ultra 253B37.6%estimated ± 11.0 pp, low confidence
121Qwen3 235B 250735.2%estimated ± 11.0 pp, low confidence
122QwQ-32B34.4%estimated ± 11.0 pp, low confidence
123Magistral Medium31.0%estimated ± 11.0 pp, low confidence
124Phi-4-reasoning29.9%estimated ± 11.0 pp, low confidence
125Magistral Small29.9%estimated ± 11.0 pp, low confidence
126LFM2.5-8B-A1B27.1%estimated ± 4.0 pp, high confidence
127LFM2.5-2.6B26.7%estimated ± 11.0 pp, low confidence
128DeepSeek V3.126.4%estimated ± 11.0 pp, low confidence
129Gemini 2.5 Flash-Lite26.4%estimated ± 11.0 pp, low confidence
130Kimi K226.4%estimated ± 11.0 pp, low confidence
131MiniCPM5-1B25.8%measured
132Claude 4 Opus25.6%estimated ± 11.0 pp, low confidence
133Claude 4 Sonnet25.6%estimated ± 11.0 pp, low confidence
134Gemini 2.0 Flash25.5%estimated ± 11.0 pp, low confidence
135Qwen3 235B-A22B25.5%estimated ± 11.0 pp, low confidence
136Gemini 2.0 Flash-Lite25.5%estimated ± 11.0 pp, low confidence
137Qwen3 32B25.4%estimated ± 11.0 pp, low confidence
138Gemini 1.5 Flash25.4%estimated ± 11.0 pp, low confidence
139LLaDA2.2-mini16.8%estimated ± 4.0 pp, medium confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General