benchgap
General

FrontierPhysics leaderboard

As of 2026-11-18, the highest measured score on FrontierPhysics is 27.4% by Claude Opus 5.5. 91 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1Claude Fable 5.250.5%estimated ± 3.2 pp, medium confidence
2Claude Opus 5.527.4%measured
3GPT-6 Astra23.2%measured
4GPT-5.5 Pro23.1%estimated ± 4.8 pp, high confidence
5GPT-6.2 Astra20.6%estimated ± 2.3 pp, high confidence
6GPT-6.2 Sol20.6%estimated ± 2.3 pp, high confidence
7GPT-6 Cyber20.6%estimated ± 2.3 pp, high confidence
8GPT-6.1 Sol19.6%measured
9Gemini 4 Carbon17.2%estimated ± 3.2 pp, high confidence
10Claude Mythos 5.215.4%estimated ± 2.3 pp, high confidence
11Claude Sonnet 5.514.8%estimated ± 3.2 pp, high confidence
12GPT-6.1 Terra14.3%measured
13GPT-6 Sol14.3%measured
14Claude Fable 5.111.3%measured
15Claude Opus 59.5%measured
16Gemini 4 Argon8.3%measured
17Claude Mythos 57.4%estimated ± 2.3 pp, high confidence
18GPT-5.4 Pro7.3%estimated ± 4.8 pp, high confidence
19GPT-5.6 Sol7.1%measured
20Kimi K3.1 Pro6.5%measured
21Muse Wonder5.7%estimated ± 3.2 pp, high confidence
22Claude Mythos Preview5.3%estimated ± 2.3 pp, high confidence
23Claude 3 Opus5.0%estimated ± 4.8 pp, medium confidence
24Claude Haiku 4.55.0%estimated ± 4.8 pp, medium confidence
25GLM-4.75.0%estimated ± 4.8 pp, medium confidence
26GLM-5.15.0%estimated ± 4.8 pp, medium confidence
27GPT-4.15.0%estimated ± 4.8 pp, medium confidence
28GPT-4.1 nano5.0%estimated ± 4.8 pp, medium confidence
29GPT-4o5.0%estimated ± 4.8 pp, medium confidence
30Grok 4.205.0%estimated ± 4.8 pp, medium confidence
31Grok 4.35.0%estimated ± 4.8 pp, medium confidence
32Inkling5.0%estimated ± 4.8 pp, medium confidence
33Kimi K2.7 Code5.0%estimated ± 4.8 pp, medium confidence
34Muse Spark 1.15.0%estimated ± 4.8 pp, high confidence
35o15.0%estimated ± 4.8 pp, medium confidence
36Seed 2.1 Pro5.0%estimated ± 4.8 pp, medium confidence
37Seed 2.1 Turbo5.0%estimated ± 4.8 pp, medium confidence
38Qwen3.8-27B4.7%estimated ± 4.5 pp, high confidence
39Claude Haiku 5.54.2%measured
40GLM-5.34.2%measured
41Grok 4.74.2%measured
42GPT-5.53.8%estimated ± 3.2 pp, high confidence
43GPT-6 Luna3.8%estimated ± 3.2 pp, high confidence
44DeepSeek V4.1 Flash 10173.6%measured
45GPT-5.6 Luna3.6%measured
46GPT-5.6 Terra3.6%measured
47Claude Opus 4.83.5%estimated ± 3.2 pp, high confidence
48Qwen3.8 Max3.4%estimated ± 4.2 pp, high confidence
49Claude Sonnet 53.1%estimated ± 4.2 pp, high confidence
50DeepSeek V4 Flash 07313.0%estimated ± 4.2 pp, high confidence
51MiniMax M3.1 Flash3.0%estimated ± 4.5 pp, high confidence
52GPT-5.22.9%estimated ± 3.2 pp, high confidence
53Gemini 3.5 Flash2.9%estimated ± 4.2 pp, high confidence
54Qwen3.7 Max2.9%estimated ± 4.2 pp, high confidence
55GPT-5.42.8%estimated ± 3.2 pp, high confidence
56o32.8%estimated ± 4.2 pp, high confidence
57Muse Spark Open2.7%estimated ± 4.5 pp, high confidence
58Hy4 preview2.7%estimated ± 4.5 pp, high confidence
59Muse Spark 1.32.7%estimated ± 4.2 pp, high confidence
60GPT-52.6%estimated ± 4.2 pp, high confidence
61Qwen3.6-35B-A3B2.6%estimated ± 4.2 pp, high confidence
62Claude Fable 52.6%estimated ± 3.2 pp, high confidence
63Gemini 3 Flash2.6%estimated ± 4.2 pp, high confidence
64Nemotron 3 Ultra2.6%estimated ± 4.2 pp, high confidence
65Qwen3.5 Flash2.6%estimated ± 4.2 pp, high confidence
66Gemini 3.5 Flash-Lite2.6%estimated ± 4.2 pp, high confidence
67Qwen 3.6 Max (preview)2.6%estimated ± 4.2 pp, high confidence
68Qwen3.5 397B2.6%estimated ± 4.2 pp, high confidence
69Claude Sonnet 4.52.6%estimated ± 4.2 pp, high confidence
70Qwen3.5-122B-A10B2.6%estimated ± 4.2 pp, high confidence
71Qwen3.5 Plus2.6%estimated ± 4.2 pp, high confidence
72Qwen3.7 Plus2.6%estimated ± 4.2 pp, high confidence
73Claude Sonnet 4.62.6%estimated ± 4.2 pp, medium confidence
74GPT-5.12.6%estimated ± 4.2 pp, medium confidence
75GLM-5.22.6%estimated ± 4.2 pp, medium confidence
76Qwen3.7 Flash2.6%estimated ± 4.2 pp, medium confidence
77GPT-4o mini2.6%estimated ± 4.2 pp, medium confidence
78Qwen3.6 Plus2.6%estimated ± 4.2 pp, medium confidence
79GLM-5.3-Flash2.6%estimated ± 4.2 pp, medium confidence
80MiniMax M32.6%estimated ± 4.2 pp, medium confidence
81GPT-4.1 mini2.6%estimated ± 4.2 pp, medium confidence
82GPT-5.4 mini2.6%estimated ± 4.2 pp, medium confidence
83o3-mini2.6%estimated ± 4.2 pp, medium confidence
84Qwen3.6-27B2.6%estimated ± 4.2 pp, medium confidence
85Inkling-Small2.6%estimated ± 4.2 pp, medium confidence
86GPT-5.4 nano2.6%estimated ± 4.2 pp, medium confidence
87o4-mini2.6%estimated ± 4.2 pp, medium confidence
88Qwen3 Max2.6%estimated ± 4.2 pp, medium confidence
89GPT-OSS 120B2.6%estimated ± 4.2 pp, medium confidence
90DeepSeek V4 Pro2.5%estimated ± 4.5 pp, high confidence
91Ling 3.1 Flash2.5%estimated ± 4.5 pp, high confidence
92Grok 4.62.4%measured
93Kimi K32.4%measured
94Claude Opus 4.72.4%estimated ± 3.2 pp, high confidence
95Step 5 Preview2.0%estimated ± 4.5 pp, high confidence
96Claude Opus 4.51.9%estimated ± 3.2 pp, high confidence
97Claude Opus 4.61.9%estimated ± 3.2 pp, high confidence
98Gemini 3.1 Pro1.7%estimated ± 3.2 pp, high confidence
99Gemini 3.7 Flash1.7%estimated ± 3.2 pp, high confidence
100Beam 1.0 501B-A23B1.6%estimated ± 4.5 pp, medium confidence
101SWE-2 Flash1.6%estimated ± 2.3 pp, high confidence
102Gemma 4 31B1.5%estimated ± 4.5 pp, medium confidence
103Gemini 3.6 Flash1.4%estimated ± 3.2 pp, high confidence
104Gemma 4 26B A4B1.4%estimated ± 4.5 pp, medium confidence
105MAI-Thinking-11.4%estimated ± 4.5 pp, medium confidence
106Grok 4.51.4%estimated ± 3.2 pp, high confidence
107Kimi K2.61.3%estimated ± 3.2 pp, high confidence
108MiMo-V2.6-Pro1.2%measured
109MiMo-V2.6-Flash1.0%estimated ± 2.3 pp, medium confidence
110Muse Spark 1.20.5%estimated ± 4.6 pp, medium confidence
111Gemini 3.8 Flash0.0%measured
112Mistral Large 40.0%measured

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General