MMMU-Pro

College-level multimodal understanding. Run by Artificial Analysis.

Published by

Artificial Analysis

Category

Multimodal

Index weight

1.0

Models

246

Data as of

9 Sept 2026

Source: Artificial Analysis (artificialanalysis.ai). Data taken from the public leaderboard.

1. 1GPT-6 Astra86.9%
2. 2GPT-6 Astra86.4%
3. 3GPT-6 Astra86.2%
4. 4Gemini 3.8 Flash85.6%
5. 5Gemini 3.7 Flash85.5%
6. 6GPT-6 Astra85.1%
7. 7Gemini 3.7 Flash84.9%
8. 8Claude Opus 584.7%
9. 9Gemini 3.7 Flash84.7%
10. 10GPT-6 Astra84.6%
11. 11Gemini 3.8 Flash84.5%
12. 12Gemini 3.5 Flash84.3%
13. 13Gemini 3.8 Flash84.2%
14. 14Claude Opus 584.0%
15. 15Gemini 3.5 Flash83.9%

SearchAll providersAlibabaAllen Institute for AIAllenaiAmazonAnthropicApodexBaiduByteDanceChina MobileCohereDeepSeekGoogleLG AI ResearchLiquid AIMetaMiniMaxMistral AIMoonshot AINVIDIAOpenAIOpenBMBSapiens AIStepFunThinking MachinesxAIXiaomiZhipu AIOpen weights246 of 246

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	GPT-6 AstraOpenAI	86.9%	GPT-6 Astra (max)	69.9	–
2	GPT-6 AstraOpenAI	86.4%	–	69.5	–
3	GPT-6 AstraOpenAI	86.2%	–	68.6	–
4	Gemini 3.8 FlashGoogle	85.6%	Gemini 3.8 Flash (high)	62.4	–
5	Gemini 3.7 FlashGoogle	85.5%	Gemini 3.7 Flash (high)	61.7	–
6	GPT-6 AstraOpenAI	85.1%	–	67.4	–
7	Gemini 3.7 FlashGoogle	84.9%	–	62.5	–
8	Claude Opus 5Anthropic	84.7%	Claude Opus 5 (max)	70.0	–
9	Gemini 3.7 FlashGoogle	84.7%	–	64.4	–
10	GPT-6 AstraOpenAI	84.6%	–	65.8	–
11	Gemini 3.8 FlashGoogle	84.5%	–	61.9	–
12	Gemini 3.5 FlashGoogle	84.3%	Gemini 3.5 Flash	60.7	–
13	Gemini 3.8 FlashGoogle	84.2%	–	63.5	–
14	Claude Opus 5Anthropic	84.0%	–	66.3	–
15	Gemini 3.5 FlashGoogle	83.9%	–	63.9	–
16	GPT-5.6 SolOpenAI	83.4%	GPT-5.6 Sol (max)	65.3	–
17	Gemini 3.6 FlashGoogle	83.2%	Gemini 3.6 Flash	60.7	–
18	GPT-5.6 SolOpenAI	82.7%	–	67.5	–
19	Claude Opus 5Anthropic	82.4%	–	69.3	–
20	Gemini 3.1 ProGoogle	82.4%	–	63.9	–
21	Qwen3 8Alibaba	82.3%	–	63.9	–
22	GPT-6 AstraOpenAI	82.0%	–	64.4	–
23	Muse Spark 1.3Meta	82.0%	–	63.9	–
24	GPT-5.6 SolOpenAI	81.8%	–	68.0	–
25	Claude Opus 5Anthropic	81.6%	–	63.1	–
26	GPT-5.6 SolOpenAI	81.4%	–	66.0	–
27	GPT-5.5OpenAI	81.2%	–	64.9	–
28	GPT-5.5OpenAI	81.1%	–	66.3	–
29	GPT-5.6 SolOpenAI	81.0%	–	64.6	–
30	GPT-5.6 TerraOpenAI	80.7%	GPT-5.6 Terra (max)	62.3	–
31	Kimi K3Moonshot AIopen	80.5%	Kimi K3 (max)	66.5	–
32	Muse SparkMeta	80.5%	–	65.4	–
33	Qwen3.7 PlusAlibaba	80.5%	Qwen3.7 Plus	59.7	–
34	Grok 4.5xAI	80.4%	Grok 4.5 (high)	63.4	–
35	Gemini 3 ProGoogle	80.2%	Gemini 3 Pro Preview (high)	61.6	–
36	Gemini 3.5 FlashGoogle	80.1%	–	57.3	–
37	Gemini 3 FlashGoogle	79.9%	–	61.5	–
38	GPT-5.5OpenAI	79.9%	GPT-5.5 (xhigh)	66.6	–
39	Claude Opus 5Anthropic	79.8%	–	61.7	–
40	Qwen3.8-Flash-NextAlibabaopen	79.8%	–	59.3	–
41	GPT-5.6 TerraOpenAI	79.5%	–	64.2	–
42	Kimi K2.6Moonshot AIopen	79.4%	Kimi K2.6	60.6	–
43	AApodex 1.1Apodex	79.2%	Apodex 1.1	–	–
44	GPT-5.6 TerraOpenAI	79.1%	–	63.1	–
45	GPT-5.5OpenAI	79.0%	–	62.3	–
46	Gemini 3.5 Flash LiteGoogle	79.0%	–	53.5	–
47	Claude Opus 4.7Anthropic	78.8%	Claude Opus 4.7 (max)	65.5	–
48	GPT-5.6 LunaOpenAI	78.5%	–	59.3	–
49	MiniMax-M3MiniMaxopen	78.5%	MiniMax-M3	56.9	–
50	Gemini 3 FlashGoogle	78.5%	–	56.6	–
51	GPT-5.6 LunaOpenAI	78.5%	GPT-5.6 Luna (max)	56.6	–
52	GPT-5.3 CodexOpenAI	78.5%	GPT-5.3 Codex (xhigh)	62.4	–
53	Kimi K3Moonshot AIopen	78.5%	–	58.8	–
54	GPT-5.4OpenAI	78.4%	GPT-5.4 (xhigh)	63.4	–
55	Grok 4.3xAI	78.1%	Grok 4.3 (high)	57.8	–
56	GPT-5.4OpenAI	78.0%	–	60.4	–
57	Qwen3.6 PlusAlibabaopen	78.0%	Qwen3.6 Plus	59.0	–
58	GPT-5.6 LunaOpenAI	77.6%	–	58.1	–
59	Claude Sonnet 5Anthropic	77.3%	Claude Sonnet 5 (max)	61.2	–
60	Qwen3.5 397B-A17BAlibabaopen	77.3%	Qwen3.5 397B A17B	58.6	–

Show 100 more