Multimodal · Benchmark profile

MMMU-Pro

A more robust extension of MMMU intended to reduce shortcut-based answering.

Ranking weightedCurrent / rollingActive54 models

Data verified 15 July 2026 · Methodology 1.2.0

Benchmark score on MMMU-Pro

Leaderboard · 54 models

Gemini 3.1 Pro Preview leads at 83.9%, followed by Gemini 3.5 Flash (83.9%) and GPT-5.6 Sol (83.4%).

Ranking · 4.2%MultimodalCurrent

#1Closed

Gemini 3.1 Pro Preview

Google

gemini-3.1-pro-preview

83.9%

Overall #42 · Context 1.05M tokens

#2Closed

Gemini 3.5 Flash

Google

gemini-3.5-flash

83.9%

Overall #17 · Context 1.05M tokens

#3Closed

GPT-5.6 Sol

OpenAI

gpt-5.6-sol

83.4%

Overall #4 · Context 1.05M tokens

Visual analysis

See the field, not just the table.

Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.

ModelsColours

GoogleOpenAIMoonshot AIxAIMetaAlibaba CloudMiniMaxAnthropic

Leaderboard bars

20 ranked models · higher is better

View accessible chart data

Accessible data for Leaderboard bars
Model	Rank	Provider	Score
Gemini 3.1 Pro Preview	#1	Google	83.9%
Gemini 3.5 Flash	#2	Google	83.9%
GPT-5.6 Sol	#3	OpenAI	83.4%
Kimi K3	#4	Moonshot AI	81.6%
GPT-5.4	#5	OpenAI	81.2%
GPT-5.5	#6	OpenAI	81.2%
Gemini 3 Pro	#7	Google	81%
GPT-5.6 Terra	#8	OpenAI	80.7%
Grok 4.5	#9	xAI	80.4%
Muse Spark	#10	Meta	80.4%
Kimi K2.6	#11	Moonshot AI	79.4%
Qwen3.7-Plus	#12	Alibaba Cloud	79%
Qwen3.6 Plus	#13	Alibaba Cloud	78.8%
GPT-5.6 Luna	#14	OpenAI	78.5%
MiniMax M3	#15	MiniMax	78.5%
GPT-5.3-Codex	#16	OpenAI	78.5%
Kimi K2.5	#17	Moonshot AI	78.5%
Grok 4.3	#18	xAI	78.1%
Claude Opus 4.6	#19	Anthropic	77.3%
Claude Sonnet 5	#20	Anthropic	77.3%

Score curve

Shows how quickly scores separate down the field

View accessible chart data

Accessible data for Score curve
Model	Rank	Provider	Score
Gemini 3.1 Pro Preview	#1	Google	83.9%
Gemini 3.5 Flash	#2	Google	83.9%
GPT-5.6 Sol	#3	OpenAI	83.4%
Kimi K3	#4	Moonshot AI	81.6%
GPT-5.4	#5	OpenAI	81.2%
GPT-5.5	#6	OpenAI	81.2%
Gemini 3 Pro	#7	Google	81%
GPT-5.6 Terra	#8	OpenAI	80.7%
Grok 4.5	#9	xAI	80.4%
Muse Spark	#10	Meta	80.4%
Kimi K2.6	#11	Moonshot AI	79.4%
Qwen3.7-Plus	#12	Alibaba Cloud	79%
Qwen3.6 Plus	#13	Alibaba Cloud	78.8%
GPT-5.6 Luna	#14	OpenAI	78.5%
MiniMax M3	#15	MiniMax	78.5%
GPT-5.3-Codex	#16	OpenAI	78.5%
Kimi K2.5	#17	Moonshot AI	78.5%
Grok 4.3	#18	xAI	78.1%
Claude Opus 4.6	#19	Anthropic	77.3%
Claude Sonnet 5	#20	Anthropic	77.3%

Provider profile

Average within the visible cohort · descriptive only

View accessible chart data

Accessible data for Provider profile
Provider	Models	Average score
Google	3	82.9%
OpenAI	6	80.6%
Meta	1	80.4%
Moonshot AI	3	79.8%
xAI	2	79.3%
Alibaba Cloud	2	78.9%
MiniMax	1	78.5%
Anthropic	2	77.3%

Benchmark score table (54 models)

One best score per model · higher is better

MMMU-Pro leaderboard ranked by score
Rank	Model	Provider	License	Score
#1	Gemini 3.1 Pro Preview gemini-3.1-pro-preview	Google	closed	83.9%
#2	Gemini 3.5 Flash gemini-3.5-flash	Google	closed	83.9%
#3	GPT-5.6 Sol gpt-5.6-sol	OpenAI	closed	83.4%
#4	Kimi K3 kimi-k3	Moonshot AI	closed	81.6%
#5	GPT-5.4 gpt-5.4	OpenAI	closed	81.2%
#6	GPT-5.5 gpt-5.5	OpenAI	closed	81.2%
#7	Gemini 3 Pro gemini-3-pro	Google	closed	81%
#8	GPT-5.6 Terra gpt-5.6-terra	OpenAI	closed	80.7%
#9	Grok 4.5 grok-4.5	xAI	closed	80.4%
#10	Muse Spark muse-spark	Meta	closed	80.4%
#11	Kimi K2.6 kimi-k2.6	Moonshot AI	open	79.4%
#12	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	79%
#13	Qwen3.6 Plus qwen3-6-plus	Alibaba Cloud	closed	78.8%
#14	GPT-5.6 Luna gpt-5.6-luna	OpenAI	closed	78.5%
#15	MiniMax M3 MiniMax-M3	MiniMax	—	78.5%
#16	GPT-5.3-Codex gpt-5.3-codex	OpenAI	closed	78.5%
#17	Kimi K2.5 kimi-k2.5	Moonshot AI	—	78.5%
#18	Grok 4.3 grok-4.3	xAI	closed	78.1%
#19	Claude Opus 4.6 claude-opus-4-6	Anthropic	closed	77.3%
#20	Claude Sonnet 5 claude-sonnet-5	Anthropic	closed	77.3%
#21	Qwen3.5 397B qwen3.5-397b-a17b	Alibaba Cloud	open	77.3%
#22	Gemma 4 31B gemma-4-31b	Google	open	76.9%
#23	GPT-5.2 Codex gpt-5.2-codex	OpenAI	closed	76.3%
#24	Gemini 3.1 Flash-Lite gemini-3.1-flash-lite	Google	closed	75.5%
#25	MiMo-V2.5 mimo-v2.5	Xiaomi	open	75.4%
#26	Qwen3.5 27B qwen3.5-27b	Alibaba Cloud	open	75.0%
#27	Qwen3.5 122B qwen3.5-122b-a10b	Alibaba Cloud	open	75.0%
#28	Gemini 2.5 Pro gemini-2.5-pro	Google	closed	74.9%
#29	Qwen3.6 27B qwen3.6-27b	Alibaba Cloud	open	74.6%
#30	Grok 4.20 grok-4.20	xAI	closed	74.6%
#31	GPT-5 gpt-5	OpenAI	closed	74.2%
#32	GPT-5 Codex gpt-5-codex	OpenAI	closed	73.8%
#33	GPT-5.4 mini gpt-5.4-mini	OpenAI	closed	73.3%
#34	Gemini 2.5 Flash gemini-2.5-flash	Google	closed	73.1%
#35	Qwen3.5 35B qwen3.5-35b-a3b	Alibaba Cloud	open	72.7%
#36	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	70.6%
#37	Qwen3.5 Omni Plus qwen3.5-omni-plus	Alibaba Cloud	closed	70.5%
#38	o3 o3	OpenAI	closed	70.1%
#39	Gemma 4 12B gemma-4-12b	Google	open	69.7%
#40	Gemma 4 26B gemma-4-26b-a4b	Google	open	69.2%
#41	o4-mini o4-mini	OpenAI	closed	69.2%
#42	GPT-5 mini gpt-5-mini	OpenAI	closed	68.8%
#43	Grok 4 grok-4	xAI	closed	68.8%
#44	Claude Sonnet 4.5 claude-sonnet-4-5	Anthropic	closed	68.7%
#45	Claude Opus 4.1 claude-opus-4-1	Anthropic	closed	67.9%
#46	GPT-5.4 nano gpt.5.4.nano	OpenAI	closed	66.1%
#47	Mistral Medium 3.5 mistral-medium-3-5	Mistral AI	open	64.9%
#48	Nova 2 Pro nova-2-pro-preview	Amazon	closed	64.5%
#49	Command A+ command-a-plus	Cohere	open	63.2%
#50	Claude Sonnet 4 claude-sonnet-4	Anthropic	closed	61.8%
#51	Grok 4 Fast grok-4-fast	xAI	closed	61.8%
#52	Mistral Small 4 mistral-small-2603	Mistral AI	open	56.8%
#53	Mistral Large 3 mistral-large-2512	Mistral AI	open	55.7%
#54	Claude Haiku 4.5 claude-haiku-4-5-20251001	Anthropic	closed	55.1%

The top of this snapshot is led by Gemini 3.1 Pro Preview at 83.9%; third place is 0.49 points behind. The top-10 spread is 3.5 points.

Compare top models

Gemini 3.1 Pro Preview vs Gemini 3.5 Flash Gemini 3.5 Flash vs GPT-5.6 Sol GPT-5.6 Sol vs Kimi K3 Kimi K3 vs GPT-5.4

About MMMU-Pro

Definition and scoring

Organisation: MMMU-Pro authors
Category: Multimodal
Version: Not evaluated
Direction: higher is better
Ranking use: Weighted 4.2%
Contamination risk: Unknown

A more robust extension of MMMU intended to reduce shortcut-based answering. Results stay tied to the exact model variant and evaluation system. Multiple systems for the same model use the best published score on this page; overall Lumina scoring uses the median of ranking-eligible rows.

Open benchmark source ↗

FAQ

Quick answers

What does MMMU-Pro measure?

A more robust extension of MMMU intended to reduce shortcut-based answering.

Which model scores highest on MMMU-Pro?

Gemini 3.1 Pro Preview by Google currently leads with 83.9%.

How many models are evaluated on MMMU-Pro?

54 models in the LuminaBench cohort have a qualifying score on this benchmark.

Does this affect overall Lumina rank?

Yes. This family is ranking-weighted at 4.2% of its capability category. Gemini 3.1 Pro Preview is overall #42.

MMMU

Rolling · 1 results · Not Weighted

MathVista

Rolling · Not evaluated · Not Weighted

ChartQA

Rolling · Not evaluated · Not Weighted

DocVQA

Rolling · Not evaluated · Not Weighted

CharXiv

May 2026 · 19 results · Weighted 4.2%

Blueprint-Bench 2

v2 · 5 results · Reference

Multimodal · Benchmark profile

MMMU-Pro

A more robust extension of MMMU intended to reduce shortcut-based answering.

Ranking weightedCurrent / rollingActive54 models

Data verified 15 July 2026 · Methodology 1.2.0

Benchmark score on MMMU-Pro

Leaderboard · 54 models

Gemini 3.1 Pro Preview leads at 83.9%, followed by Gemini 3.5 Flash (83.9%) and GPT-5.6 Sol (83.4%).

Ranking · 4.2%MultimodalCurrent

#1Closed

Gemini 3.1 Pro Preview

Google

gemini-3.1-pro-preview

83.9%

Overall #42 · Context 1.05M tokens

#2Closed

Gemini 3.5 Flash

Google

gemini-3.5-flash

83.9%

Overall #17 · Context 1.05M tokens

#3Closed

GPT-5.6 Sol

OpenAI

gpt-5.6-sol

83.4%

Overall #4 · Context 1.05M tokens

Visual analysis

See the field, not just the table.

Three views of the same sourced leaderboard: model placement, the score curve, and descriptive provider averages. These visuals never alter ranking calculations.

ModelsColours

GoogleOpenAIMoonshot AIxAIMetaAlibaba CloudMiniMaxAnthropic

Leaderboard bars

20 ranked models · higher is better

View accessible chart data

Accessible data for Leaderboard bars
Model	Rank	Provider	Score
Gemini 3.1 Pro Preview	#1	Google	83.9%
Gemini 3.5 Flash	#2	Google	83.9%
GPT-5.6 Sol	#3	OpenAI	83.4%
Kimi K3	#4	Moonshot AI	81.6%
GPT-5.4	#5	OpenAI	81.2%
GPT-5.5	#6	OpenAI	81.2%
Gemini 3 Pro	#7	Google	81%
GPT-5.6 Terra	#8	OpenAI	80.7%
Grok 4.5	#9	xAI	80.4%
Muse Spark	#10	Meta	80.4%
Kimi K2.6	#11	Moonshot AI	79.4%
Qwen3.7-Plus	#12	Alibaba Cloud	79%
Qwen3.6 Plus	#13	Alibaba Cloud	78.8%
GPT-5.6 Luna	#14	OpenAI	78.5%
MiniMax M3	#15	MiniMax	78.5%
GPT-5.3-Codex	#16	OpenAI	78.5%
Kimi K2.5	#17	Moonshot AI	78.5%
Grok 4.3	#18	xAI	78.1%
Claude Opus 4.6	#19	Anthropic	77.3%
Claude Sonnet 5	#20	Anthropic	77.3%

Score curve

Shows how quickly scores separate down the field

View accessible chart data

Accessible data for Score curve
Model	Rank	Provider	Score
Gemini 3.1 Pro Preview	#1	Google	83.9%
Gemini 3.5 Flash	#2	Google	83.9%
GPT-5.6 Sol	#3	OpenAI	83.4%
Kimi K3	#4	Moonshot AI	81.6%
GPT-5.4	#5	OpenAI	81.2%
GPT-5.5	#6	OpenAI	81.2%
Gemini 3 Pro	#7	Google	81%
GPT-5.6 Terra	#8	OpenAI	80.7%
Grok 4.5	#9	xAI	80.4%
Muse Spark	#10	Meta	80.4%
Kimi K2.6	#11	Moonshot AI	79.4%
Qwen3.7-Plus	#12	Alibaba Cloud	79%
Qwen3.6 Plus	#13	Alibaba Cloud	78.8%
GPT-5.6 Luna	#14	OpenAI	78.5%
MiniMax M3	#15	MiniMax	78.5%
GPT-5.3-Codex	#16	OpenAI	78.5%
Kimi K2.5	#17	Moonshot AI	78.5%
Grok 4.3	#18	xAI	78.1%
Claude Opus 4.6	#19	Anthropic	77.3%
Claude Sonnet 5	#20	Anthropic	77.3%

Provider profile

Average within the visible cohort · descriptive only

View accessible chart data

Accessible data for Provider profile
Provider	Models	Average score
Google	3	82.9%
OpenAI	6	80.6%
Meta	1	80.4%
Moonshot AI	3	79.8%
xAI	2	79.3%
Alibaba Cloud	2	78.9%
MiniMax	1	78.5%
Anthropic	2	77.3%

Benchmark score table (54 models)

One best score per model · higher is better

MMMU-Pro leaderboard ranked by score
Rank	Model	Provider	License	Score
#1	Gemini 3.1 Pro Preview gemini-3.1-pro-preview	Google	closed	83.9%
#2	Gemini 3.5 Flash gemini-3.5-flash	Google	closed	83.9%
#3	GPT-5.6 Sol gpt-5.6-sol	OpenAI	closed	83.4%
#4	Kimi K3 kimi-k3	Moonshot AI	closed	81.6%
#5	GPT-5.4 gpt-5.4	OpenAI	closed	81.2%
#6	GPT-5.5 gpt-5.5	OpenAI	closed	81.2%
#7	Gemini 3 Pro gemini-3-pro	Google	closed	81%
#8	GPT-5.6 Terra gpt-5.6-terra	OpenAI	closed	80.7%
#9	Grok 4.5 grok-4.5	xAI	closed	80.4%
#10	Muse Spark muse-spark	Meta	closed	80.4%
#11	Kimi K2.6 kimi-k2.6	Moonshot AI	open	79.4%
#12	Qwen3.7-Plus qwen3.7-plus	Alibaba Cloud	—	79%
#13	Qwen3.6 Plus qwen3-6-plus	Alibaba Cloud	closed	78.8%
#14	GPT-5.6 Luna gpt-5.6-luna	OpenAI	closed	78.5%
#15	MiniMax M3 MiniMax-M3	MiniMax	—	78.5%
#16	GPT-5.3-Codex gpt-5.3-codex	OpenAI	closed	78.5%
#17	Kimi K2.5 kimi-k2.5	Moonshot AI	—	78.5%
#18	Grok 4.3 grok-4.3	xAI	closed	78.1%
#19	Claude Opus 4.6 claude-opus-4-6	Anthropic	closed	77.3%
#20	Claude Sonnet 5 claude-sonnet-5	Anthropic	closed	77.3%
#21	Qwen3.5 397B qwen3.5-397b-a17b	Alibaba Cloud	open	77.3%
#22	Gemma 4 31B gemma-4-31b	Google	open	76.9%
#23	GPT-5.2 Codex gpt-5.2-codex	OpenAI	closed	76.3%
#24	Gemini 3.1 Flash-Lite gemini-3.1-flash-lite	Google	closed	75.5%
#25	MiMo-V2.5 mimo-v2.5	Xiaomi	open	75.4%
#26	Qwen3.5 27B qwen3.5-27b	Alibaba Cloud	open	75.0%
#27	Qwen3.5 122B qwen3.5-122b-a10b	Alibaba Cloud	open	75.0%
#28	Gemini 2.5 Pro gemini-2.5-pro	Google	closed	74.9%
#29	Qwen3.6 27B qwen3.6-27b	Alibaba Cloud	open	74.6%
#30	Grok 4.20 grok-4.20	xAI	closed	74.6%
#31	GPT-5 gpt-5	OpenAI	closed	74.2%
#32	GPT-5 Codex gpt-5-codex	OpenAI	closed	73.8%
#33	GPT-5.4 mini gpt-5.4-mini	OpenAI	closed	73.3%
#34	Gemini 2.5 Flash gemini-2.5-flash	Google	closed	73.1%
#35	Qwen3.5 35B qwen3.5-35b-a3b	Alibaba Cloud	open	72.7%
#36	Claude Opus 4.5 claude-opus-4-5	Anthropic	closed	70.6%
#37	Qwen3.5 Omni Plus qwen3.5-omni-plus	Alibaba Cloud	closed	70.5%
#38	o3 o3	OpenAI	closed	70.1%
#39	Gemma 4 12B gemma-4-12b	Google	open	69.7%
#40	Gemma 4 26B gemma-4-26b-a4b	Google	open	69.2%
#41	o4-mini o4-mini	OpenAI	closed	69.2%
#42	GPT-5 mini gpt-5-mini	OpenAI	closed	68.8%
#43	Grok 4 grok-4	xAI	closed	68.8%
#44	Claude Sonnet 4.5 claude-sonnet-4-5	Anthropic	closed	68.7%
#45	Claude Opus 4.1 claude-opus-4-1	Anthropic	closed	67.9%
#46	GPT-5.4 nano gpt.5.4.nano	OpenAI	closed	66.1%
#47	Mistral Medium 3.5 mistral-medium-3-5	Mistral AI	open	64.9%
#48	Nova 2 Pro nova-2-pro-preview	Amazon	closed	64.5%
#49	Command A+ command-a-plus	Cohere	open	63.2%
#50	Claude Sonnet 4 claude-sonnet-4	Anthropic	closed	61.8%
#51	Grok 4 Fast grok-4-fast	xAI	closed	61.8%
#52	Mistral Small 4 mistral-small-2603	Mistral AI	open	56.8%
#53	Mistral Large 3 mistral-large-2512	Mistral AI	open	55.7%
#54	Claude Haiku 4.5 claude-haiku-4-5-20251001	Anthropic	closed	55.1%

The top of this snapshot is led by Gemini 3.1 Pro Preview at 83.9%; third place is 0.49 points behind. The top-10 spread is 3.5 points.

Compare top models

Gemini 3.1 Pro Preview vs Gemini 3.5 Flash Gemini 3.5 Flash vs GPT-5.6 Sol GPT-5.6 Sol vs Kimi K3 Kimi K3 vs GPT-5.4

About MMMU-Pro

Definition and scoring

Organisation: MMMU-Pro authors
Category: Multimodal
Version: Not evaluated
Direction: higher is better
Ranking use: Weighted 4.2%
Contamination risk: Unknown

Open benchmark source ↗

FAQ

Quick answers

What does MMMU-Pro measure?

A more robust extension of MMMU intended to reduce shortcut-based answering.

Which model scores highest on MMMU-Pro?

Gemini 3.1 Pro Preview by Google currently leads with 83.9%.

How many models are evaluated on MMMU-Pro?

54 models in the LuminaBench cohort have a qualifying score on this benchmark.

Does this affect overall Lumina rank?

Yes. This family is ranking-weighted at 4.2% of its capability category. Gemini 3.1 Pro Preview is overall #42.

MMMU

Rolling · 1 results · Not Weighted

MathVista

Rolling · Not evaluated · Not Weighted

ChartQA

Rolling · Not evaluated · Not Weighted

DocVQA

Rolling · Not evaluated · Not Weighted

CharXiv

May 2026 · 19 results · Weighted 4.2%

Blueprint-Bench 2

v2 · 5 results · Reference

Leaderboard · 54 models

Gemini 3.1 Pro Preview

Gemini 3.5 Flash

GPT-5.6 Sol

See the field, not just the table.

Benchmark score table (54 models)

Compare top models

Definition and scoring

Quick answers

What does MMMU-Pro measure?

Which model scores highest on MMMU-Pro?

How many models are evaluated on MMMU-Pro?

Does this affect overall Lumina rank?

More Multimodal benchmarks

Leaderboard · 54 models

Gemini 3.1 Pro Preview

Gemini 3.5 Flash

GPT-5.6 Sol

See the field, not just the table.

Benchmark score table (54 models)

Compare top models

Definition and scoring

Quick answers

What does MMMU-Pro measure?

Which model scores highest on MMMU-Pro?

How many models are evaluated on MMMU-Pro?

Does this affect overall Lumina rank?

More Multimodal benchmarks