Vision & documents

MMMU-Pro w/ Python leaderboard

Name: benchgap: LLM benchmark scores, measured and estimated
Creator: benchgap

As of 2026-10-07, the highest measured score on MMMU-Pro w/ Python is 84.6% by GPT-5.6 Sol. 105 more models have estimated scores, calibrated from the benchmarks they were measured on.

Measured scores: benchlm.ai.

#	Model	Score	Source
1	Claude Opus 5.5	86.2%	estimated ± 1.6 pp, medium confidence
2	GPT-6 Astra	85.8%	estimated ± 1.6 pp, medium confidence
3	Gemini 3.1 Pro	85.3%	estimated ± 0.6 pp, medium confidence
4	GPT-6.1 Sol	85.3%	estimated ± 1.6 pp, medium confidence
5	Gemini 3.8 Flash	85.0%	estimated ± 1.6 pp, medium confidence
6	Gemini 3.5 Flash	85.0%	estimated ± 0.6 pp, medium confidence
7	Gemini 3.7 Flash	85.0%	estimated ± 1.6 pp, medium confidence
8	GPT-5.6 Sol	84.6%	measured
9	Claude Opus 5	84.5%	estimated ± 1.6 pp, medium confidence
10	Qwen3.8 Max	83.7%	estimated ± 0.6 pp, high confidence
11	Gemini 3.6 Flash	83.7%	estimated ± 1.6 pp, high confidence
12	GPT-6 Sol	83.5%	estimated ± 1.6 pp, high confidence
13	Qwen3.8 Max Preview	83.4%	estimated ± 1.6 pp, high confidence
14	Kimi K3	83.4%	measured
15	GPT-5.5	83.2%	measured
16	Seed 2.1 Pro	83.0%	estimated ± 0.6 pp, high confidence
17	Gemini 3 Pro	82.4%	estimated ± 0.6 pp, high confidence
18	GPT-5.4	82.1%	measured
19	GPT-5.6 Terra	82.0%	measured
20	Grok 4.5	82.0%	estimated ± 1.6 pp, high confidence
21	Muse Spark	81.7%	estimated ± 0.6 pp, high confidence
22	Qwen3.8-Flash-Next	81.6%	estimated ± 1.6 pp, high confidence
23	GPT-6 Luna	81.6%	estimated ± 1.6 pp, high confidence
24	Seed 2.1 Turbo	81.4%	estimated ± 0.6 pp, high confidence
25	Apodex 1.1	81.3%	estimated ± 1.6 pp, high confidence
26	Apodex 1.1 Mini	81.3%	estimated ± 1.6 pp, high confidence
27	Gemini 3.5 Flash-Lite	81.1%	estimated ± 1.6 pp, high confidence
28	Ling 3.0 Flash VL	81.1%	estimated ± 1.6 pp, high confidence
29	Claude Opus 4.7 (Adaptive)	81.0%	estimated ± 1.6 pp, high confidence
30	Gemini 3 Flash	80.9%	estimated ± 1.6 pp, high confidence
31	GPT-5.2	80.8%	estimated ± 0.6 pp, high confidence
32	GPT-5.3 Codex	80.8%	estimated ± 1.6 pp, high confidence
33	dots3-note Preview	80.4%	estimated ± 0.6 pp, high confidence
34	Qwen3.5 397B	80.3%	estimated ± 0.6 pp, high confidence
35	Qwen3.7 Plus	80.3%	estimated ± 0.6 pp, high confidence
36	Qwen3.6 Plus	80.1%	estimated ± 0.6 pp, high confidence
37	Kimi K2.6	80.1%	measured
38	Claude Sonnet 5	80.1%	estimated ± 1.6 pp, high confidence
39	DeepSeek V4.1 Flash	79.9%	estimated ± 1.6 pp, high confidence
40	Kimi K2.5	79.8%	estimated ± 0.6 pp, high confidence
41	Kimi K2.5 (Reasoning)	79.8%	estimated ± 0.6 pp, high confidence
42	Claude Opus 4.7	79.5%	estimated ± 1.6 pp, high confidence
43	Mistral Large 4	79.5%	estimated ± 1.6 pp, high confidence
44	GPT-5.6 Luna	79.5%	measured
45	GPT-5.2-Codex	79.5%	estimated ± 1.6 pp, high confidence
46	Qwen3.8-27B	79.5%	estimated ± 1.6 pp, high confidence
47	Grok 4.3	79.5%	estimated ± 0.6 pp, high confidence
48	MiniMax M3	79.5%	estimated ± 0.6 pp, high confidence
49	Pareto 26.9	79.4%	estimated ± 0.6 pp, high confidence
50	MiMo-V2.5	79.3%	estimated ± 0.6 pp, high confidence
51	GPT-5.1	79.0%	estimated ± 1.6 pp, high confidence
52	Claude Opus 4.6 (Adaptive)	78.9%	estimated ± 1.6 pp, high confidence
53	Step 3.7 Flash	78.8%	estimated ± 1.6 pp, high confidence
54	Claude Opus 4.6	78.7%	estimated ± 0.6 pp, high confidence
55	Qwen3.5-122B-A10B	78.7%	estimated ± 1.6 pp, high confidence
56	Qwen3.5-27B	78.7%	estimated ± 1.6 pp, high confidence
57	Gemini 2.5 Pro	78.6%	estimated ± 1.6 pp, high confidence
58	Gemma 4 31B	78.3%	estimated ± 0.6 pp, high confidence
59	GPT-5 (medium)	78.2%	estimated ± 1.6 pp, high confidence
60	GPT-5 (high)	78.1%	estimated ± 1.6 pp, high confidence
61	Claude Opus 4.5 Thinking	78.0%	estimated ± 1.6 pp, high confidence
62	GPT-5.4 mini	78.0%	measured
63	Step 5 Preview	77.4%	estimated ± 0.6 pp, high confidence
64	MiMo-V2.6-Flash	77.4%	estimated ± 1.6 pp, high confidence
65	Qwen3.6-27B	77.2%	estimated ± 0.6 pp, high confidence
66	GLM-5V-Turbo	77.2%	estimated ± 1.6 pp, high confidence
67	Qwen3.5-35B-A3B	77.2%	estimated ± 1.6 pp, high confidence
68	GPT-5.1-Codex	77.0%	estimated ± 1.6 pp, high confidence
69	GPT-5.1-Codex-Max	77.0%	estimated ± 1.6 pp, high confidence
70	Qwen3.6-35B-A3B	76.8%	estimated ± 0.6 pp, high confidence
71	Grok 4.20	76.7%	estimated ± 0.6 pp, high confidence
72	Claude Sonnet 4.6	75.8%	estimated ± 1.6 pp, high confidence
73	Inkling-Small	75.6%	estimated ± 0.6 pp, high confidence
74	Muse Glimmer 30B	75.6%	estimated ± 0.6 pp, high confidence
75	o3	75.5%	estimated ± 1.6 pp, high confidence
76	Gemma 4 26B A4B	75.4%	estimated ± 0.6 pp, high confidence
77	MiMo-V2-Omni	75.3%	estimated ± 1.6 pp, high confidence
78	Inkling	75.1%	estimated ± 0.6 pp, high confidence
79	Grok 4	74.6%	estimated ± 1.6 pp, high confidence
80	Claude 4.1 Opus Thinking	74.0%	estimated ± 1.6 pp, high confidence
81	Interfaze Beta	73.1%	estimated ± 0.6 pp, high confidence
82	Claude Opus 4.5	72.7%	estimated ± 0.6 pp, high confidence
83	Gemini 2.5 Flash	72.3%	estimated ± 1.6 pp, high confidence
84	Mistral Medium 3.5 128B	71.9%	estimated ± 1.6 pp, medium confidence
85	Gemma 4 12B	71.5%	estimated ± 0.6 pp, high confidence
86	Grok 4.1 Fast (Reasoning)	70.7%	estimated ± 1.6 pp, medium confidence
87	Claude 4 Sonnet	70.1%	estimated ± 1.6 pp, medium confidence
88	Llama 4 Maverick	69.8%	estimated ± 1.6 pp, medium confidence
89	Grok 4 Fast (Reasoning)	69.6%	estimated ± 1.6 pp, medium confidence
90	GPT-5.4 nano	69.5%	measured
91	GPT-4.1	69.2%	estimated ± 1.6 pp, medium confidence
92	Qwen3-Omni-30B-A3B-Thinking	68.4%	estimated ± 1.6 pp, medium confidence
93	Command A+	67.5%	estimated ± 0.6 pp, medium confidence
94	GPT-4.1 mini	67.3%	estimated ± 1.6 pp, medium confidence
95	Mistral Small 4	65.8%	estimated ± 1.6 pp, medium confidence
96	Mistral Small 4 (Reasoning)	65.8%	estimated ± 1.6 pp, medium confidence
97	Mistral Large 3	65.0%	estimated ± 1.6 pp, medium confidence
98	Qwen3-Omni-30B-A3B-Instruct	64.8%	estimated ± 1.6 pp, medium confidence
99	Gemini 1.5 Pro	64.4%	estimated ± 1.6 pp, medium confidence
100	Nemotron 3 Nano Omni 30B A3B	63.0%	estimated ± 1.6 pp, medium confidence
101	Mistral Medium 3	62.8%	estimated ± 1.6 pp, medium confidence
102	Llama 4 Scout	62.7%	estimated ± 1.6 pp, medium confidence
103	Qwen3.5 397B (Reasoning)	62.6%	estimated ± 1.6 pp, medium confidence
104	Gemma 4 E4B	61.5%	estimated ± 1.6 pp, medium confidence
105	LFM2.5-VL-3B	60.9%	estimated ± 0.6 pp, medium confidence
106	Grok 4.1 Fast	59.0%	estimated ± 1.6 pp, medium confidence
107	Gemma 3 27B	58.6%	estimated ± 1.6 pp, medium confidence
108	Gemma 4 E2B	55.6%	estimated ± 1.6 pp, medium confidence
109	Nova Pro	55.4%	estimated ± 1.6 pp, medium confidence
110	GPT-4o mini	52.8%	estimated ± 1.6 pp, medium confidence
111	GPT-4.1 nano	51.5%	estimated ± 1.6 pp, medium confidence
112	Claude 3 Haiku	42.0%	estimated ± 1.6 pp, medium confidence
113	LFM2.5-VL-1.6B-Extract	37.3%	estimated ± 1.6 pp, medium confidence
114	Phi-4 Multimodal Instruct	22.3%	estimated ± 1.6 pp, medium confidence

MMMU-Pro w/ Python leaderboard

All leaderboards

Agentic · terminal

Agentic · tools

Coding

Math

Knowledge & reasoning

Instruction following

Multilingual

Vision & documents

Long context

General