benchgap
Vision & documents

MedXpertQA (MM) leaderboard

As of 2026-11-21, the highest measured score on MedXpertQA (MM) is 81.3% by Gemini 3.1 Pro. 111 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.1 Terra85.2%estimated ± 2.4 pp, medium confidence
2Kimi K3.1 Pro82.7%estimated ± 2.4 pp, medium confidence
3Gemini 3.5 Flash81.7%estimated ± 2.4 pp, high confidence
4Gemini 3.1 Pro81.3%measured
5GPT-5.6 Sol80.6%estimated ± 2.4 pp, high confidence
6Claude Mythos 580.4%estimated ± 6.4 pp, low confidence
7Qwen3.8 Max80.4%measured
8Glimpse80.3%estimated ± 4.2 pp, low confidence
9Claude Opus 4.779.4%estimated ± 6.4 pp, low confidence
10LongCat-2.579.1%estimated ± 6.4 pp, low confidence
11Claude Opus 4.879.0%estimated ± 6.4 pp, low confidence
12GLM-5.3-Flash78.8%estimated ± 6.4 pp, low confidence
13Qwen3.8-Flash-Next78.5%estimated ± 4.2 pp, medium confidence
14Gemini 3.7 Flash78.5%estimated ± 6.4 pp, low confidence
15Muse Spark78.4%measured
16Muse Spark 1.178.4%estimated ± 6.4 pp, low confidence
17Claude Sonnet 578.4%estimated ± 6.4 pp, low confidence
18Qwen3.8-Omni-Flash78.2%estimated ± 4.2 pp, medium confidence
19GPT-6.2 Astra77.9%estimated ± 6.5 pp, low confidence
20Inkling 2 Small77.9%estimated ± 2.4 pp, high confidence
21Kimi K377.9%estimated ± 2.4 pp, high confidence
22Seed 2.1 Pro77.9%estimated ± 2.4 pp, high confidence
23Claude Opus 5.577.9%estimated ± 6.5 pp, low confidence
24GPT-6 Astra77.8%estimated ± 6.5 pp, low confidence
25GPT-6.1 Sol77.8%estimated ± 6.5 pp, low confidence
26Gemini 3.8 Flash77.8%estimated ± 6.5 pp, low confidence
27Claude Opus 577.8%estimated ± 6.5 pp, low confidence
28Sakana Fugu-Ultra77.7%estimated ± 6.4 pp, low confidence
29Gemini 3.6 Flash77.6%estimated ± 6.5 pp, low confidence
30GPT-6 Sol77.6%estimated ± 6.5 pp, low confidence
31Qwen3.8 Max Preview77.6%estimated ± 6.5 pp, low confidence
32Muse Spark Open77.2%estimated ± 6.5 pp, low confidence
33GPT-5.477.1%measured
34Grok 4.577.1%estimated ± 6.5 pp, low confidence
35Sakana Fugu77.1%estimated ± 6.4 pp, low confidence
36GPT-5.577.1%estimated ± 2.4 pp, high confidence
37GPT-6 Luna76.9%estimated ± 6.5 pp, low confidence
38Gemini 3 Pro76.6%estimated ± 2.4 pp, high confidence
39Apodex 1.176.6%estimated ± 6.5 pp, low confidence
40Apodex 1.1 Mini76.6%estimated ± 6.5 pp, low confidence
41Gemini 3.5 Flash-Lite76.5%estimated ± 6.5 pp, low confidence
42Ling 3.0 Flash VL76.5%estimated ± 6.5 pp, low confidence
43Gemini 3 Flash76.3%estimated ± 6.5 pp, low confidence
44Qwen3.8-27B76.3%estimated ± 4.2 pp, medium confidence
45GPT-5.3 Codex76.3%estimated ± 6.5 pp, low confidence
46GPT-5.6 Terra76.0%estimated ± 2.4 pp, high confidence
47MiniMax M3.1 Flash76.0%estimated ± 6.4 pp, low confidence
48Nemotron 3.5 Super VL 120B A12B75.3%estimated ± 6.4 pp, low confidence
49DeepSeek V4.1 Flash 101775.0%estimated ± 6.5 pp, low confidence
50Nemotron 3.5 Super VL 120B A12B NVFP475.0%estimated ± 6.4 pp, low confidence
51Seed 2.1 Turbo74.8%estimated ± 2.4 pp, high confidence
52GPT-5.2-Codex74.2%estimated ± 6.5 pp, low confidence
53Claude Sonnet 4.674.0%estimated ± 6.4 pp, low confidence
54Nemotron 3 Nano Omni 30B A3B73.5%estimated ± 6.4 pp, low confidence
55GPT-5.273.5%estimated ± 2.4 pp, high confidence
56Kimi K2.673.3%estimated ± 2.4 pp, high confidence
57GPT-5.173.0%estimated ± 6.5 pp, low confidence
58Claude Opus 4.6 (Adaptive)72.8%estimated ± 6.5 pp, low confidence
59Step 3.7 Flash72.6%estimated ± 6.5 pp, low confidence
60dots3-note Preview72.6%estimated ± 2.4 pp, high confidence
61Qwen3.5 397B72.4%estimated ± 2.4 pp, high confidence
62Gemini 3.1 Flash-Lite72.3%estimated ± 6.4 pp, low confidence
63Qwen3.6 Plus72.0%estimated ± 2.4 pp, high confidence
64Gemini 2.5 Pro71.9%estimated ± 6.5 pp, low confidence
65Hy471.5%estimated ± 2.4 pp, high confidence
66Kimi K2.571.3%estimated ± 2.4 pp, high confidence
67GPT-5.6 Luna71.1%estimated ± 2.4 pp, high confidence
68Qwen3.7 Plus71.0%measured
69Grok 4.370.4%estimated ± 2.4 pp, high confidence
70MiniMax M370.4%estimated ± 2.4 pp, high confidence
71GPT-570.2%estimated ± 6.5 pp, low confidence
72Pareto 26.970.2%estimated ± 2.4 pp, high confidence
73MiMo-V2.570.0%estimated ± 2.4 pp, high confidence
74Claude Opus 4.5 Thinking69.7%estimated ± 6.5 pp, low confidence
75Gemma 4 31B67.7%estimated ± 2.4 pp, high confidence
76Qwen3.5-122B-A10B67.7%estimated ± 2.4 pp, high confidence
77GPT-5.4 mini67.0%estimated ± 2.4 pp, high confidence
78MiMo-V2.6-Flash66.8%estimated ± 6.5 pp, low confidence
79Grok 4.2065.8%measured
80GLM-5V-Turbo65.7%estimated ± 6.5 pp, low confidence
81Step 5 Preview65.6%estimated ± 2.4 pp, high confidence
82Hy4 flash65.1%estimated ± 2.4 pp, high confidence
83Qwen3.6-27B65.1%estimated ± 2.4 pp, high confidence
84LFM2.5-VL-3B65.0%estimated ± 7.1 pp, low confidence
85Claude Opus 4.664.8%measured
86GPT-5.1-Codex64.5%estimated ± 6.5 pp, low confidence
87GPT-5.1-Codex-Max64.5%estimated ± 6.5 pp, low confidence
88Qwen3.6-35B-A3B64.0%estimated ± 2.4 pp, high confidence
89Qwen3.5-35B-A3B63.5%estimated ± 2.4 pp, high confidence
90Qwen3.5-27B63.2%estimated ± 2.4 pp, high confidence
91Inkling 2 Mini63.0%estimated ± 2.4 pp, high confidence
92Inkling-Small60.8%estimated ± 2.4 pp, high confidence
93Muse Glimmer 30B60.8%estimated ± 2.4 pp, high confidence
94Gemma 4 26B A4B60.3%estimated ± 2.4 pp, high confidence
95Inkling59.6%estimated ± 2.4 pp, high confidence
96LFM3-VL-2B58.1%estimated ± 4.2 pp, low confidence
97Interfaze Beta53.7%estimated ± 2.4 pp, high confidence
98Claude Opus 4.552.4%estimated ± 2.4 pp, high confidence
99o351.5%estimated ± 6.5 pp, low confidence
100MiMo-V2-Omni50.2%estimated ± 6.5 pp, low confidence
101Gemma 4 12B48.7%measured
102Qwen3-VL-235B-A22B Instruct46.1%estimated ± 2.4 pp, medium confidence
103o4-mini45.3%estimated ± 6.5 pp, low confidence
104Grok 442.5%estimated ± 6.5 pp, low confidence
105GPT-5.4 nano41.2%estimated ± 2.4 pp, medium confidence
106GLM-4.6V40.9%estimated ± 2.4 pp, medium confidence
107Command A+33.8%estimated ± 2.4 pp, medium confidence
108GLM-4.6V-Flash28.4%estimated ± 2.4 pp, medium confidence
109Gemini 2.5 Flash20.1%estimated ± 6.5 pp, low confidence
110Qwen3.5-4B19.5%estimated ± 6.5 pp, low confidence
111Mistral Medium 3.5 128B16.9%estimated ± 6.5 pp, low confidence
112Grok 4.1 Fast (Reasoning)10.3%estimated ± 6.5 pp, low confidence
113Claude 4 Sonnet7.6%estimated ± 6.5 pp, low confidence
114Llama 4 Maverick6.9%estimated ± 6.5 pp, low confidence
115Grok 4 Fast (Reasoning)6.2%estimated ± 6.5 pp, low confidence
116GPT-4.15.0%estimated ± 6.5 pp, low confidence
117Qwen3-Omni-30B-A3B-Thinking3.5%estimated ± 6.5 pp, low confidence
118Claude 3.7 Sonnet3.4%estimated ± 6.5 pp, low confidence
119GPT-4.1 mini2.0%estimated ± 6.5 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General