BenchLeader

MMMU-Pro leaderboard

As of 15 Nov 2026, the top score on MMMU-Pro is 89.3% by Gemini 4 Carbon (Google), ahead of Gemini 4 Argon at 88.3%, across 296 model configurations with a published result (Artificial Analysis, result dated 28 Oct 2026). MMMU-Pro is a harder version of the college-level multimodal exam MMMU: questions with images from science, engineering, medicine and the arts, with ten answer options and a vision-only variant where the question itself is inside the image. MMMU-Pro counts toward the BenchLeader multimodal score.

Key facts (as of 15 Nov 2026)

  • Gemini 4 Carbon holds the top MMMU-Pro score at 89.3% (Artificial Analysis, 28 Oct 2026).
  • Gemini 4 Argon is second on MMMU-Pro at 88.3% (Artificial Analysis, 30 Sep 2026).
  • MMMU-Pro: College-level multimodal understanding. Run by Artificial Analysis (Artificial Analysis).
  • 296 model configurations have a published MMMU-Pro result on BenchLeader (BenchLeader, 15 Nov 2026).
  • MMMU-Pro feeds the multimodal category of the BenchLeader Index (BenchLeader methodology).

College-level multimodal understanding. Run by Artificial Analysis.

Published by
Artificial Analysis
Category
Multimodal
Index weight
1.0
Models
296
Data as of
15 Nov 2026

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

What the test looks like

MMMU-Pro is a harder version of the college-level multimodal exam MMMU: questions with images from science, engineering, medicine and the arts, with ten answer options and a vision-only variant where the question itself is inside the image.

How it is scored

Percent correct; Artificial Analysis runs the standard setting itself.

What to keep in mind

Requires vision, so text-only models are absent. Ten options make guessing less effective than on the original MMMU, so scores are lower.

296 of 296
#
1Gemini 4 Carbonmedium89.3%65.82026-10-28
2Gemini 4 Argonmedium88.3%61.62026-09-30
3Claude Opus 5.5max87.7%68.42026-09-22
4GPT-6 Astramax86.9%67.32026-09-03
5Claude Opus 5.5xhigh86.6%65.42026-09-22
6GPT-6 Astrahigh86.4%66.32026-09-03
7GPT-6 Astraxhigh86.2%66.02026-09-03
8GPT-6.1 Solmax86.0%66.62026-09-29
9Claude Opus 5.5high85.8%66.02026-09-22
10Claude Opus 5.5medium85.7%61.22026-09-22
11Gemini 3.8 Flashhigh85.6%59.42026-09-02
12Gemini 3.7 Flashhigh85.5%62.12026-08-13
13GPT-6 Astramedium85.1%64.42026-09-03
14GPT-6.1 Solxhigh85.1%63.62026-09-29
15GPT-6.1 Solhigh84.9%63.42026-09-29
16Gemini 3.7 Flashlow84.9%57.32026-08-13
17Claude Opus 5max84.7%65.02026-07-24
18Gemini 3.7 Flashmedium84.7%60.12026-08-13
19GPT-6.1 Terramax84.7%64.12026-10-26
20Claude Opus 5.5low84.7%58.92026-09-22
21GPT-6 Astralow84.6%62.02026-09-03
22Gemini 3.8 Flashlow84.5%55.92026-09-02
23Gemini 3.5 Flashhigh84.3%62.52026-05-19
24Gemini 3.8 Flashmedium84.2%60.22026-09-02
25Claude Opus 5xhigh84.0%60.12026-07-24
26GPT-6.1 Solmedium83.9%62.72026-09-29
27Kimi K3.1 Pro83.9%61.82026-10-23
28Gemini 3.5 Flashmedium83.9%61.82026-05-19
29GPT-6.1 Terraxhigh83.8%60.02026-10-26
30GPT-6.1 Terrahigh83.8%59.62026-10-26
31Muse Wonderhigh83.8%56.22026-10-12
32GPT-5.6 Solmax83.4%65.32026-07-09
33Gemini 3.6 Flashhigh83.2%61.02026-07-21
34GPT-6.1 Sollow83.1%60.02026-09-29
35GPT-6 Solmax83.0%62.62026-09-22
36Qwen3.8 Maxmax82.8%58.72026-09-02
37GPT-6.1 Terramedium82.8%58.52026-10-26
38GPT-5.6 Solxhigh82.7%63.92026-07-09
39GPT-6 Solxhigh82.5%58.42026-09-22
40GPT-6.1 Terralow82.5%56.62026-10-26
41Claude Opus 5high82.4%64.42026-07-24
42Gemini 3.1 Pro82.4%59.92026-02-19
43Muse Spark 1.3xhigh82.0%58.12026-09-02
44GPT-6 Solhigh82.0%58.12026-09-22
45GPT-5.6 Solhigh81.8%63.42026-07-09
46Claude Opus 5medium81.6%57.52026-07-24
47Inkling 2 Small81.6%57.92026-11-05
48GPT-5.6 Solmedium81.4%61.32026-07-09
49GPT-5.5medium81.2%60.82026-04-23
50GPT-5.5high81.1%63.52026-04-23
51GPT-5.6 Sollow81.0%60.22026-07-09
52Muse Spark Openxhigh80.9%53.52026-10-15
53GPT-5.6 Terramax80.7%59.52026-07-09
54DeepSeek V4.1 Flash 1017max80.6%57.72026-10-17
55Muse Spark80.5%64.22026-04-08
56Kimi K3max80.5%60.22026-07-16
57Qwen3.7 Plus80.5%57.62026-06-01
58GPT-6 Solmedium80.4%56.92026-09-22
59Grok 4.5high80.4%56.82026-07-08
60Gemini 3 Prohigh80.2%60.02025-11-18

Multimodal AI benchmark leaderboard · All LLM benchmarks

Cite as: BenchLeader, “MMMU-Pro leaderboard”, data as of 15 Nov 2026, https://www.benchleader.com/benchmarks/aa_mmmu_pro.

MMMU-Pro: questions

What does MMMU-Pro measure?
MMMU-Pro is a harder version of the college-level multimodal exam MMMU: questions with images from science, engineering, medicine and the arts, with ten answer options and a vision-only variant where the question itself is inside the image. Scores are reported in percent of tasks solved; higher is better.
What is the top score on MMMU-Pro?
As of 15 Nov 2026, the highest score on MMMU-Pro is 89.3% by Gemini 4 Carbon (result dated 28 Oct 2026), ahead of Gemini 4 Argon at 88.3%. Scores come from Artificial Analysis and are re-read daily.
Which AI model leads MMMU-Pro?
Gemini 4 Carbon leads MMMU-Pro with 89.3% as of 15 Nov 2026 (result dated 28 Oct 2026), ahead of Gemini 4 Argon at 88.3%.
How many models have MMMU-Pro results?
296 model configurations have a MMMU-Pro result on BenchLeader, all taken from Artificial Analysis.
Who runs MMMU-Pro and how often is it updated?
MMMU-Pro is published by Artificial Analysis. BenchLeader re-reads the published results once a day and records the date each result was published.
Does MMMU-Pro count toward the BenchLeader Index?
Yes. MMMU-Pro contributes to the multimodal category of the BenchLeader Index, normalised so that 50 is the average of the evaluated models.