benchgap
Knowledge & reasoning

HLE-Verified leaderboard

As of 2026-11-21, the highest measured score on HLE-Verified is 54.9% by Gemini 3.8 Flash. 149 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.290.9%estimated ± 7.1 pp, low confidence
2Gemini 4 Neon60.2%estimated ± 7.1 pp, low confidence
3GPT-6.2 Astra55.7%estimated ± 1.3 pp, low confidence
4GPT-6 Astra55.6%estimated ± 1.3 pp, low confidence
5GPT-6.1 Sol55.4%estimated ± 1.3 pp, low confidence
6Gemini 4 Carbon55.0%estimated ± 1.3 pp, medium confidence
7Gemini 3.8 Flash54.9%measured
8Claude Opus 5.554.9%estimated ± 1.3 pp, medium confidence
9Gemini 4 Argon54.8%estimated ± 1.2 pp, low confidence
10Grok 4.654.5%estimated ± 2.0 pp, medium confidence
11GPT-5.6 Sol54.5%measured
12Claude Fable 5.154.5%estimated ± 1.3 pp, medium confidence
13Claude Opus 554.4%measured
14Claude Sonnet 5.554.4%estimated ± 6.8 pp, low confidence
15Muse Wonder Full54.4%estimated ± 6.8 pp, low confidence
16Kimi K3.1 Pro54.4%estimated ± 6.8 pp, low confidence
17Muse Wonder54.4%estimated ± 6.8 pp, low confidence
18MiMo-V2.6-Pro54.4%estimated ± 6.8 pp, low confidence
19MiniMax M3.1 Flash54.3%estimated ± 6.8 pp, low confidence
20GPT-6 Sol54.3%estimated ± 1.3 pp, medium confidence
21Step 5 Preview54.3%estimated ± 6.8 pp, low confidence
22GPT-6.1 Terra54.3%estimated ± 6.8 pp, low confidence
23Qwen3.8 Max54.3%estimated ± 4.6 pp, medium confidence
24Claude Fable 554.2%estimated ± 1.3 pp, medium confidence
25Kimi K354.1%estimated ± 2.0 pp, medium confidence
26Muse Spark 1.354.1%estimated ± 2.0 pp, medium confidence
27Claude Haiku 5.554.0%estimated ± 6.8 pp, low confidence
28Gemini 3.7 Flash53.6%measured
29Grok 4.553.5%estimated ± 2.0 pp, medium confidence
30GPT-6.2 Sol53.2%estimated ± 1.3 pp, medium confidence
31MiniMax M353.0%estimated ± 2.0 pp, medium confidence
32GPT-5.552.9%estimated ± 1.3 pp, medium confidence
33DeepSeek V4 Pro52.7%estimated ± 2.0 pp, medium confidence
34Gemini 3.6 Flash52.7%estimated ± 2.0 pp, medium confidence
35Qwen3.8 Max Preview52.7%estimated ± 2.0 pp, medium confidence
36GPT-5.5 Pro52.6%estimated ± 1.3 pp, medium confidence
37GPT-5.4 Pro52.2%estimated ± 1.3 pp, medium confidence
38Grok 4.751.9%estimated ± 6.8 pp, low confidence
39Inkling 2 Small51.5%estimated ± 1.3 pp, low confidence
40dots3-note Preview51.4%estimated ± 1.3 pp, low confidence
41GPT-5.6 Terra51.1%measured
42Qwen3.7 Max49.9%estimated ± 2.0 pp, medium confidence
43Qwen3.8-Flash-Next49.9%estimated ± 2.0 pp, medium confidence
44Gemini 3.1 Pro49.0%estimated ± 1.3 pp, low confidence
45Gemini 3.5 Flash49.0%estimated ± 2.0 pp, medium confidence
46Claude Opus 4.748.2%estimated ± 1.3 pp, low confidence
47Gemini 4 Hydrogen46.9%estimated ± 7.1 pp, low confidence
48GPT-5.446.9%estimated ± 1.3 pp, low confidence
49Claude Opus 4.846.8%estimated ± 2.0 pp, medium confidence
50GLM-5.342.3%estimated ± 2.0 pp, medium confidence
51Claude Opus 4.641.3%estimated ± 7.1 pp, low confidence
52Qwen4 Max38.8%estimated ± 7.1 pp, low confidence
53GPT-5.3 Codex38.7%estimated ± 2.0 pp, medium confidence
54Claude Opus 4.535.7%estimated ± 7.1 pp, low confidence
55GPT-5.1-Codex34.6%estimated ± 7.1 pp, low confidence
56GPT-5.1-Codex-Max34.6%estimated ± 7.1 pp, low confidence
57Gemini 2.5 Pro33.8%estimated ± 7.1 pp, low confidence
58Ling 3.0 Tiny32.5%estimated ± 7.1 pp, low confidence
59Claude Sonnet 531.0%measured
60GPT-5.6 Luna31.0%estimated ± 2.0 pp, medium confidence
61Kimi K2.631.0%estimated ± 2.0 pp, medium confidence
62o129.0%estimated ± 7.1 pp, low confidence
63Grok 4.2026.0%estimated ± 4.6 pp, low confidence
64DeepSeek V4 Flash 073125.8%estimated ± 2.0 pp, low confidence
65Gemini 3 Pro25.8%estimated ± 2.0 pp, low confidence
66DeepSeek-R125.0%estimated ± 7.1 pp, low confidence
67GLM-4.723.8%estimated ± 7.1 pp, low confidence
68GLM-5V-Turbo23.8%estimated ± 7.1 pp, low confidence
69Claude 3.7 Sonnet22.8%estimated ± 7.1 pp, low confidence
70Claude 3.7 Sonnet Thinking22.6%estimated ± 7.1 pp, low confidence
71Muse Glimmer 30B22.6%estimated ± 7.1 pp, low confidence
72GPT-4.122.4%estimated ± 7.1 pp, low confidence
73Kimi K222.1%estimated ± 7.1 pp, low confidence
74Qwen3.8-27B21.9%estimated ± 2.0 pp, low confidence
75Muse Spark 1.220.8%estimated ± 2.0 pp, low confidence
76Gemini 2.5 Flash20.8%estimated ± 7.1 pp, low confidence
77Mistral Large 420.8%estimated ± 7.1 pp, low confidence
78Step 3.7 Flash20.6%estimated ± 7.1 pp, low confidence
79DeepSeek V320.3%estimated ± 7.1 pp, low confidence
80GPT-5.220.0%estimated ± 2.0 pp, low confidence
81Mistral Large 319.8%estimated ± 7.1 pp, low confidence
82o4-mini19.7%estimated ± 7.1 pp, low confidence
83Qwen3 Max19.3%estimated ± 7.1 pp, low confidence
84DeepSeek V3 032419.2%estimated ± 7.1 pp, low confidence
85DeepSeek V3.218.9%estimated ± 7.1 pp, low confidence
86Grok 4.318.6%estimated ± 2.0 pp, low confidence
87Gemini 2.0 Flash18.5%estimated ± 7.1 pp, low confidence
88Grok Code Fast 118.5%estimated ± 7.1 pp, low confidence
89DeepSeek V3.118.1%estimated ± 7.1 pp, low confidence
90Mistral Medium 3.5 128B18.1%estimated ± 7.1 pp, low confidence
91Qwen3.7 Plus18.0%estimated ± 2.0 pp, low confidence
92Claude 4 Sonnet17.8%estimated ± 7.1 pp, low confidence
93GPT-5.2-Codex17.6%estimated ± 2.0 pp, low confidence
94Mercury 2.517.2%estimated ± 7.1 pp, low confidence
95Muse Spark 1.117.2%estimated ± 2.0 pp, low confidence
96Nemotron 3.5 Bolt 74B-A9B BF1617.0%estimated ± 7.1 pp, low confidence
97GPT-OSS 120B16.9%estimated ± 7.1 pp, low confidence
98Hy316.8%estimated ± 2.0 pp, low confidence
99Hy3 Preview16.8%estimated ± 2.0 pp, low confidence
100GLM-4.616.6%estimated ± 7.1 pp, low confidence
101Mistral Small 416.6%estimated ± 7.1 pp, low confidence
102Claude Opus 4.6 (Adaptive)16.6%estimated ± 2.0 pp, low confidence
103Kimi K2.7 Code16.6%estimated ± 2.0 pp, low confidence
104Muse Ember16.4%estimated ± 7.1 pp, low confidence
105GLM-5.216.4%estimated ± 2.0 pp, low confidence
106Inkling-Small16.4%estimated ± 2.0 pp, low confidence
107Qwen3.5-27B16.0%estimated ± 7.1 pp, low confidence
108Qwen3.5 397B16.0%estimated ± 2.0 pp, low confidence
109Inkling 2 Mini15.9%estimated ± 7.1 pp, low confidence
110Solar Pro 415.8%estimated ± 2.0 pp, low confidence
111Muse Spark Open15.7%estimated ± 2.0 pp, low confidence
112GPT-4.1 mini15.7%estimated ± 7.1 pp, low confidence
113Magistral Medium15.7%estimated ± 7.1 pp, low confidence
114Qwen 3.6 Max (preview)15.6%estimated ± 2.0 pp, low confidence
115Qwen3.5-35B-A3B15.5%estimated ± 7.1 pp, low confidence
116Muse Spark15.4%estimated ± 2.0 pp, low confidence
117Qwen3.6 Plus15.4%estimated ± 2.0 pp, low confidence
118Kimi K2.515.4%estimated ± 2.0 pp, low confidence
119Gemma 4 31B15.3%estimated ± 7.1 pp, low confidence
120GPT-4o15.3%estimated ± 7.1 pp, low confidence
121Mistral Large 215.3%estimated ± 7.1 pp, low confidence
122Grok 415.3%estimated ± 2.0 pp, low confidence
123GPT-5.4 mini15.3%estimated ± 2.0 pp, low confidence
124MiniMax M2.715.3%estimated ± 2.0 pp, low confidence
125GPT-5.115.3%estimated ± 2.0 pp, low confidence
126Inkling15.3%estimated ± 2.0 pp, low confidence
127MiMo-V2-Pro15.3%estimated ± 2.0 pp, low confidence
128Gemini 3 Flash15.3%estimated ± 4.6 pp, low confidence
129Qwen3.6-27B15.1%estimated ± 7.1 pp, low confidence
130Ling 3.1 Flash3.6%estimated ± 6.8 pp, low confidence
131GLM-5.3-Flash3.2%estimated ± 4.6 pp, low confidence
132DeepSeek V4.1 Flash 10172.7%estimated ± 6.8 pp, low confidence
133Nemotron 3 Ultra2.3%estimated ± 4.6 pp, low confidence
134Claude Opus 4.5 Thinking1.8%estimated ± 4.6 pp, low confidence
135Claude Sonnet 4.61.2%estimated ± 4.6 pp, low confidence
136GPT-51.2%estimated ± 4.6 pp, low confidence
137Grok 4 Fast (Reasoning)1.0%estimated ± 4.6 pp, low confidence
138GPT-6 Luna0.9%estimated ± 6.8 pp, low confidence
139Ling 3.0 Flash0.5%estimated ± 4.6 pp, low confidence
140GLM-5.10.3%estimated ± 4.6 pp, low confidence
141Grok 4.1 Fast (Reasoning)0.3%estimated ± 4.6 pp, low confidence
142o30.2%estimated ± 4.6 pp, low confidence
143Gemini 3.5 Flash-Lite0.1%estimated ± 4.6 pp, low confidence
144Qwen3.5 Flash0.0%estimated ± 4.6 pp, low confidence
145MiMo-V2.5-Pro0.0%estimated ± 6.8 pp, low confidence
146MiMo-V2.6-Flash0.0%estimated ± 6.8 pp, low confidence
147Beam 1.0 501B-A23B0.0%estimated ± 6.8 pp, low confidence
148Apodex 1.10.0%estimated ± 6.8 pp, low confidence
149Apodex 1.1 Mini0.0%estimated ± 6.8 pp, low confidence
150Nemotron 3.5 Super VL 120B A12B0.0%estimated ± 6.8 pp, low confidence
151Nemotron 3.5 Super VL 120B A12B NVFP40.0%estimated ± 6.8 pp, low confidence
152A.X K20.0%estimated ± 6.8 pp, low confidence
153GLM-50.0%estimated ± 6.8 pp, low confidence
154GPT-5.4 nano0.0%estimated ± 6.8 pp, low confidence
155GLM-5-Turbo0.0%estimated ± 6.8 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General