benchgap
Knowledge & reasoning

MedXpertQA (Text) leaderboard

As of 2026-11-22, the highest measured score on MedXpertQA (Text) is 71.5% by Gemini 3.1 Pro. 77 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.2 Astra83.5%estimated ± 4.5 pp, low confidence
2GPT-6.2 Sol77.7%estimated ± 4.5 pp, low confidence
3GPT-6 Astra77.7%estimated ± 4.5 pp, low confidence
4Sakana Fugu75.1%estimated ± 4.5 pp, low confidence
5Sakana Fugu-Ultra75.1%estimated ± 4.5 pp, low confidence
6Claude Mythos 574.6%estimated ± 8.9 pp, low confidence
7Claude Sonnet 5.572.9%estimated ± 8.9 pp, low confidence
8Claude Opus 572.4%estimated ± 8.9 pp, low confidence
9Gemini 3.1 Pro71.5%measured
10GPT-5.6 Sol70.7%estimated ± 4.5 pp, low confidence
11Muse Spark 1.168.9%estimated ± 8.9 pp, low confidence
12Claude Opus 4.768.9%estimated ± 4.5 pp, medium confidence
13Kimi K3.1 Pro68.9%estimated ± 4.5 pp, medium confidence
14Qwen4 Max68.5%estimated ± 8.9 pp, low confidence
15Claude Opus 4.866.3%estimated ± 4.5 pp, medium confidence
16GPT-5.566.3%estimated ± 4.5 pp, medium confidence
17Pareto 26.966.1%estimated ± 8.9 pp, low confidence
18Kimi K365.9%estimated ± 4.5 pp, medium confidence
19Step 5 Preview65.9%estimated ± 4.5 pp, medium confidence
20Inkling 2 Small63.9%estimated ± 4.5 pp, medium confidence
21ERNIE 5.5 Flash Preview63.7%estimated ± 8.9 pp, low confidence
22GPT-5.6 Terra63.5%estimated ± 4.5 pp, medium confidence
23Claude Haiku 5.563.2%estimated ± 8.9 pp, low confidence
24Ornith-1.5-397B63.2%estimated ± 4.5 pp, medium confidence
25Gemini 3.5 Flash62.7%estimated ± 4.5 pp, medium confidence
26Qwen3.8 Max62.4%estimated ± 4.5 pp, medium confidence
27Pareto 26.10 Preview61.7%estimated ± 4.5 pp, medium confidence
28Qwen3.7 Max61.7%estimated ± 4.5 pp, medium confidence
29GPT-5.6 Luna61.3%estimated ± 4.5 pp, medium confidence
30Hy4 preview61.3%estimated ± 4.5 pp, medium confidence
31MiniMax M3.1 Flash60.8%estimated ± 8.9 pp, low confidence
32Claude Sonnet 560.7%estimated ± 8.9 pp, low confidence
33GPT-5.5 Pro60.6%estimated ± 8.9 pp, low confidence
34GPT-5.4 Pro60.2%estimated ± 8.9 pp, low confidence
35GPT-5.459.6%measured
36Qwen3.8-Flash-Next59.2%estimated ± 4.5 pp, medium confidence
37GLM-5.257.5%estimated ± 4.5 pp, medium confidence
38Qwen3.8-Omni-Flash56.8%estimated ± 4.5 pp, medium confidence
39DeepSeek V4.1 Flash 101756.5%estimated ± 4.5 pp, medium confidence
40Beam 1.0 501B-A23B55.6%estimated ± 4.5 pp, medium confidence
41Kimi K2.655.3%estimated ± 4.5 pp, medium confidence
42Particle-155.0%estimated ± 4.5 pp, medium confidence
43Qwen3.7 Plus54.7%estimated ± 4.5 pp, medium confidence
44Beam 1.0 501B-A23B FP854.7%estimated ± 4.5 pp, medium confidence
45DeepSeek V4 Pro54.1%estimated ± 4.5 pp, medium confidence
46Interfaze Beta53.5%estimated ± 4.5 pp, medium confidence
47Nemotron 3.5 Super VL 120B A12B NVFP453.5%estimated ± 4.5 pp, medium confidence
48Muse Spark52.6%measured
49Nemotron 3.5 Super VL 120B A12B52.4%estimated ± 4.5 pp, medium confidence
50Inkling-Small52.3%estimated ± 4.5 pp, medium confidence
51Claude Opus 4.652.1%measured
52Beam 1.0 501B-A23B NVFP451.8%estimated ± 4.5 pp, medium confidence
53Mercury 351.6%estimated ± 4.5 pp, medium confidence
54Ornith-1.5-35B-A3B51.5%estimated ± 4.5 pp, medium confidence
55Qwen3.8-27B51.5%estimated ± 4.5 pp, medium confidence
56MiMo-V2.5-Pro51.0%estimated ± 8.9 pp, low confidence
57Ling 3.0 Flash Sante51.0%measured
58Solar Pro 450.9%estimated ± 4.5 pp, medium confidence
59Nemotron 3.5 Bolt 74B-A9B BF1650.5%estimated ± 4.5 pp, medium confidence
60Grok 4.2050.2%measured
61OrcaCyber-Nano-1.049.0%estimated ± 4.5 pp, low confidence
62DeepSeek V4 Flash 073148.6%estimated ± 4.5 pp, low confidence
63Inkling48.1%estimated ± 4.5 pp, low confidence
64Kimi K2.547.4%estimated ± 4.5 pp, low confidence
65Hy3 Preview46.4%estimated ± 4.5 pp, low confidence
66MiniMax M2.745.9%estimated ± 4.5 pp, low confidence
67Nemotron 3 Ultra45.9%estimated ± 4.5 pp, low confidence
68Inkling 2 Mini45.0%estimated ± 4.5 pp, low confidence
69Nemotron 3.5 Bolt 74B-A9B44.7%estimated ± 4.5 pp, low confidence
70Ornith-1.5-9B44.6%estimated ± 4.5 pp, low confidence
71Solar Open 244.4%estimated ± 4.5 pp, low confidence
72GPT-5.4 mini44.3%estimated ± 8.9 pp, low confidence
73GLM-5.144.2%estimated ± 4.5 pp, low confidence
74GLM-543.7%estimated ± 4.5 pp, low confidence
75Ternary Bonsai 2 27B43.2%estimated ± 4.5 pp, low confidence
76A.X K242.9%estimated ± 4.5 pp, low confidence
77Ripple-65B Preview42.1%estimated ± 4.5 pp, low confidence
78Ling 3.0 Flash41.6%estimated ± 4.5 pp, low confidence
79MAI-Thinking-140.1%estimated ± 4.5 pp, low confidence
80Ling 3.0 Flash FP839.8%estimated ± 4.5 pp, low confidence
81GPT-5.4 nano39.3%estimated ± 8.9 pp, low confidence
82Gemma 4 31B32.8%estimated ± 8.9 pp, low confidence
83Spark-X2.6-4B30.6%estimated ± 8.9 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General