MMMU (validation)

Official MMMU validation-set leaderboard.

Published by

MMMU

Category

Multimodal

Index weight

0.5

Models

46

Data as of

9 Sept 2026

MMMU (Yue et al.).

1. 1GPT-5.185.4%
2. 2Gemini 2.5 Deep Think84.0%
3. 3o382.9%
4. 4o4-mini81.6%
5. 5Claude Opus 4.580.7%
6. 6Gemini 2.5 Flash79.7%
7. 7Gemini 2.5 Pro79.6%
8. 8Qwen3-VL 235B-A22B78.7%
9. 9o178.2%
10. 10Grok 378.0%
11. 11Claude Sonnet 4.577.8%
12. 12Claude Opus 4.177.1%
13. 13Claude Opus 476.5%
14. 14Claude 3.7 Sonnet75.0%
15. 15Seed 1.674.8%

SearchAll providersAlibabaAmazonAnthropicByteDanceGoogleMetaMicrosoftNVIDIAOpenAIOpenBMB / LLaVARekaxAIOpen weights46 of 46

#	Model	Score ↓	Run ↕	Index ↕	Published ↕
1	GPT-5.1OpenAI	85.4%	GPT-5.1	59.4	–
2	Gemini 2.5 Deep ThinkGoogle	84.0%	–	–	–
3	o3OpenAI	82.9%	o3	60.3	–
4	o4-miniOpenAI	81.6%	–	54.9	–
5	Claude Opus 4.5Anthropic	80.7%	Claude Opus 4.5	58.0	–
6	Gemini 2.5 FlashGoogle	79.7%	–	48.8	–
7	Gemini 2.5 ProGoogle	79.6%	–	54.1	–
8	Qwen3-VL 235B-A22BAlibabaopen	78.7%	–	49.6	–
9	o1OpenAI	78.2%	o1	53.2	–
10	Grok 3xAI	78.0%	–	50.0	–
11	Claude Sonnet 4.5Anthropic	77.8%	Claude Sonnet 4.5	52.3	–
12	Claude Opus 4.1Anthropic	77.1%	Claude Opus 4.1	52.6	–
13	Claude Opus 4Anthropic	76.5%	Claude Opus 4	50.7	–
14	Claude 3.7 SonnetAnthropic	75.0%	Claude 3.7 Sonnet	49.8	–
15	Seed 1.6ByteDance	74.8%	–	–	–
16	GPT-4.5OpenAI	74.4%	GPT‑4.5	49.6	–
17	Claude Sonnet 4Anthropic	74.4%	Claude Sonnet 4	49.5	–
18	Llama 4 MaverickMetaopen	73.4%	Llama 4 Maverick	42.6	–
19	Gemini 2.0 ProGoogle	72.7%	Gemini 2.0 Pro	47.0	–
20	Gemini 2.0 FlashGoogle	71.7%	Gemini 2.0 Flash	42.3	–
21	GPT-4oOpenAI	70.7%	1120	42.3	–
22	Llama 4 ScoutMetaopen	69.4%	Llama 4 Scout	39.7	–
23	Claude 3.5 SonnetAnthropic	68.3%	Claude 3.5 Sonnet	45.1	–
24	Gemini 2.0 Flash-LiteGoogle	68.0%	–	47.0	–
25	Nova PremierAmazon	68.0%	–	45.3	–
26	Gemini 1.5 ProGoogle	65.8%	0801	42.9	–
27	Gemma 3 27BGoogleopen	64.9%	–	39.7	–
28	Qwen2 Vl 72BAlibaba	64.5%	–	–	–
29	Qwen2.5-VL 72B InstructAlibabaopen	64.5%	–	–	–
30	Nova ProAmazon	62.0%	–	41.0	–
31	Llama 3.2 90BMeta	60.3%	Llama 3.2 90B	–	–
32	Gemma 3 12B ITNVIDIAopen	59.6%	–	37.2	–
33	Claude 3 OpusAnthropic	59.4%	Claude 3 Opus	38.9	–
34	GPT-4o miniOpenAI	59.4%	GPT-4o mini	35.8	–
35	Qwen2.5-VL 7B InstructAlibabaopen	58.6%	–	–	–
36	RReka CoreReka	56.3%	–	–	–
37	Gemini 1.5 FlashGoogle	56.1%	Gemini 1.5 Flash	–	–
38	RReka FlashReka	53.3%	–	43.1	–
39	Claude 3 SonnetAnthropic	53.1%	Claude 3 Sonnet	37.9	–
40	Qwen VlAlibaba	51.4%	–	–	–
41	Claude 3 HaikuAnthropic	50.2%	Claude 3 Haiku	36.9	–
42	OMinicpm V 2.6OpenBMB / LLaVA	49.8%	MiniCPM-V 2.6	–	–
43	Gemma 3 4B ITNVIDIAopen	48.8%	–	34.5	–
44	Gemini 1.0 ProGoogle	47.9%	Gemini 1.0 Pro	–	–
45	Qwen-VL PlusAlibaba	45.2%	–	–	–
46	Phi 3.5 VisionMicrosoft	43.0%	–	–	–