benchgap
General

MaxxBench leaderboard

As of 2026-11-18, the highest measured score on MaxxBench is 76.4% by Claude Opus 5.5. 93 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#ModelScoreSource
1GPT-6.2 Astra76.8%estimated ± 4.3 pp, low confidence
2Claude Fable 5.276.6%estimated ± 4.3 pp, low confidence
3Claude Opus 5.576.4%measured
4Gemini 4 Carbon76.2%estimated ± 4.3 pp, low confidence
5GPT-6.1 Terra74.7%estimated ± 4.3 pp, low confidence
6Gemini 4 Argon72.1%estimated ± 4.3 pp, medium confidence
7GPT-6 Astra71.9%measured
8Claude Fable 5.170.2%estimated ± 4.3 pp, medium confidence
9GPT-6.1 Sol69.4%estimated ± 4.3 pp, medium confidence
10GPT-6 Sol69.1%estimated ± 4.3 pp, medium confidence
11GPT-5.6 Sol69.1%measured
12Claude Sonnet 5.566.9%measured
13GPT-5.5 Pro66.7%estimated ± 10.3 pp, low confidence
14Claude Haiku 5.565.7%estimated ± 4.3 pp, medium confidence
15GPT-5.4 Pro64.9%estimated ± 10.3 pp, low confidence
16Qwen3.8 Max64.8%estimated ± 5.7 pp, medium confidence
17Claude Fable 563.5%measured
18Claude Opus 563.5%measured
19GPT-5.561.1%estimated ± 4.3 pp, low confidence
20Kimi K3.1 Pro61.0%estimated ± 5.7 pp, medium confidence
21GPT-5.6 Terra59.9%estimated ± 5.6 pp, medium confidence
22Muse Wonder59.3%estimated ± 4.3 pp, low confidence
23Grok 4.658.7%estimated ± 4.3 pp, low confidence
24Qwen3.8-27B57.6%estimated ± 5.7 pp, medium confidence
25Claude Opus 4.857.4%estimated ± 4.3 pp, low confidence
26Kimi K357.0%estimated ± 5.7 pp, medium confidence
27GLM-5.356.7%measured
28MiMo-V2.6-Flash56.2%estimated ± 5.7 pp, medium confidence
29DeepSeek V4.1 Flash 101755.6%measured
30MiMo-V2.6-Pro55.2%estimated ± 5.7 pp, medium confidence
31GLM-5.3-Flash55.1%measured
32GPT-5.454.9%estimated ± 4.3 pp, low confidence
33Grok 4.754.8%estimated ± 5.7 pp, medium confidence
34Claude Sonnet 553.8%estimated ± 5.6 pp, medium confidence
35MiniMax M3.1 Flash53.6%estimated ± 5.7 pp, medium confidence
36Gemini 3.8 Flash53.4%measured
37Muse Spark Open52.7%estimated ± 5.7 pp, medium confidence
38GPT-5.252.7%estimated ± 4.3 pp, low confidence
39Hy4 preview52.6%estimated ± 5.7 pp, medium confidence
40Mistral Large 452.2%estimated ± 5.7 pp, medium confidence
41DeepSeek V4 Pro51.9%estimated ± 5.7 pp, medium confidence
42GPT-6 Luna51.7%measured
43Ling 3.1 Flash51.6%estimated ± 5.7 pp, medium confidence
44Gemini 3.7 Flash50.7%estimated ± 6.6 pp, low confidence
45Muse Spark 1.349.7%estimated ± 5.7 pp, medium confidence
46Gemini 3.6 Flash49.4%estimated ± 6.6 pp, low confidence
47Step 5 Preview49.3%estimated ± 5.7 pp, medium confidence
48Muse Spark 1.149.1%estimated ± 10.3 pp, low confidence
49MiniMax M349.1%estimated ± 5.7 pp, medium confidence
50Grok 4.549.1%estimated ± 6.6 pp, low confidence
51Kimi K2.648.8%estimated ± 6.6 pp, low confidence
52Claude Opus 4.748.4%estimated ± 4.3 pp, low confidence
53o347.8%estimated ± 8.8 pp, medium confidence
54Qwen3.5 Flash47.3%estimated ± 8.8 pp, medium confidence
55Qwen 3.6 Max (preview)47.2%estimated ± 8.8 pp, medium confidence
56Qwen3.5 397B47.2%estimated ± 8.8 pp, medium confidence
57Claude Sonnet 4.547.2%estimated ± 8.8 pp, medium confidence
58Qwen3.5-122B-A10B47.2%estimated ± 8.8 pp, medium confidence
59Qwen3.5 Plus47.2%estimated ± 8.8 pp, medium confidence
60Qwen3.7 Plus47.2%estimated ± 8.8 pp, medium confidence
61Claude Sonnet 4.647.2%estimated ± 8.8 pp, medium confidence
62GPT-5.147.2%estimated ± 8.8 pp, medium confidence
63Qwen3.7 Flash47.2%estimated ± 8.8 pp, medium confidence
64GPT-4o mini47.2%estimated ± 8.8 pp, medium confidence
65Qwen3.6 Plus47.2%estimated ± 8.8 pp, medium confidence
66GPT-4.1 mini47.2%estimated ± 8.8 pp, medium confidence
67GPT-5.4 mini47.2%estimated ± 8.8 pp, medium confidence
68o3-mini47.2%estimated ± 8.8 pp, medium confidence
69GPT-5.4 nano47.2%estimated ± 8.8 pp, low confidence
70o4-mini47.2%estimated ± 8.8 pp, low confidence
71Qwen3 Max47.2%estimated ± 8.8 pp, low confidence
72GPT-OSS 120B47.2%estimated ± 8.8 pp, low confidence
73GPT-5.6 Luna46.1%measured
74Grok 4.2046.0%estimated ± 10.3 pp, low confidence
75Qwen3.6-27B45.0%estimated ± 5.7 pp, low confidence
76Beam 1.0 501B-A23B44.6%estimated ± 5.7 pp, low confidence
77Seed 2.1 Turbo44.1%estimated ± 10.3 pp, low confidence
78Gemma 4 31B43.1%estimated ± 5.7 pp, low confidence
79Claude Opus 4.542.0%estimated ± 4.3 pp, low confidence
80Gemini 3.1 Pro42.0%estimated ± 4.3 pp, low confidence
81Inkling-Small41.7%estimated ± 5.7 pp, low confidence
82GLM-4.741.0%estimated ± 10.3 pp, low confidence
83Inkling41.0%estimated ± 10.3 pp, low confidence
84Seed 2.1 Pro41.0%estimated ± 10.3 pp, low confidence
85GLM-5.140.3%estimated ± 10.3 pp, low confidence
86Claude Opus 4.639.3%estimated ± 4.3 pp, low confidence
87GPT-539.3%estimated ± 4.3 pp, low confidence
88Grok 4.339.1%estimated ± 10.3 pp, low confidence
89Muse Spark 1.238.8%measured
90Hy338.8%estimated ± 11.1 pp, low confidence
91Gemma 4 26B A4B37.6%estimated ± 5.7 pp, low confidence
92Qwen3.6-35B-A3B37.3%estimated ± 5.7 pp, low confidence
93Kimi K2.7 Code36.6%estimated ± 10.3 pp, low confidence
94GPT-4.134.7%estimated ± 10.3 pp, low confidence
95Nemotron 3 Ultra34.6%estimated ± 5.7 pp, low confidence
96o134.1%estimated ± 10.3 pp, low confidence
97GPT-4o33.4%estimated ± 10.3 pp, low confidence
98GLM-5.233.1%estimated ± 4.3 pp, low confidence
99Qwen3.7 Max33.1%estimated ± 4.3 pp, low confidence
100MAI-Thinking-132.9%estimated ± 5.7 pp, low confidence
101Claude Haiku 4.532.8%estimated ± 10.3 pp, low confidence
102Claude 3 Opus30.3%estimated ± 10.3 pp, low confidence
103DeepSeek V4 Flash 073128.1%measured
104GPT-4.1 nano27.8%estimated ± 10.3 pp, low confidence
105Gemini 3.5 Flash20.4%estimated ± 4.3 pp, low confidence
106Gemini 3.5 Flash-Lite0.0%estimated ± 5.6 pp, low confidence
107Gemini 3 Flash0.0%estimated ± 5.6 pp, low confidence

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General