Skip to content

Benchmark

MMMU-Pro

Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.

Models and systems measured
77
Scores recorded
135
Sources
3
IQ weight
2.51

Score distribution

73.5%
Mean
76.9%
Median
11.3%
Std dev
59.9%
10th percentile
82.3%
90th percentile
32.7%
Min
32.7%90%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 2.51 for chat models, measured over 75 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#Model or systemScore
1Kimi K3.1 Pro· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
90%
2GPT-6.1 Sol· LLM Stats data86%
3GPT-5.6 Solwith tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.84.6%
4Gemini 3.5 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.6%
5Gemini 3.5 FlashGemini 3.5 Flash Model Card · May 2026 · no tools· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.6%
6Kimi K3Python tool · three-run average · max effort· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.4%
7GPT 5.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83.2%
8GPT 5.5with tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.2%
9GPT 5.5with tools · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83.2%
10GPT-5.6 Sol· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim83%
11GPT-5.6 Solno tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.83%
12Doubao Seed 2.1 Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82.7%
13Qwen3.8 Max· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82.3%
14Qwen3.8 MaxQwen3.8-Max release vision/video table · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.3%
15Doubao Seed 2.1 Turbo· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim82.2%
16GPT 5.4with tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82.1%
17GPT-5.6 Terrawith tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.82%
18Inkling 2 Small· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.6%
19Kimi K3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.6%
20Kimi K3no tools · three-run average · max effort · temp 1.0 · top_p .95· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.6%
21Gemini 3 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.2%
22Gemini 3 FlashGemini 3.5 Flash Model Card · May 2026 · no tools· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
23Gemini 3 Flashno tools · Standard and Vision average· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
24GPT 5.4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81.2%
25GPT 5.4no tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
26GPT 5.5Gemini 3.5 Flash Model Card · May 2026 · no tools· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
27GPT 5.5no tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
28GPT 5.5no tools · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81.2%
29Gemini 3 Pro Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim81%
30Gemini 3 Pro PreviewGemini 3 Pro Model Card · results as of November 2025· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81%
31Gemini 3 Pro PreviewOpus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted · no tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.81%
32GPT-5.6 Terra· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.7%
33GPT-5.6 Terrano tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.7%
34Gemini 3.1 Pro Preview· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.5%
35Gemini 3.1 Pro PreviewGemini 3.5 Flash Model Card · May 2026 · no tools· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.5%
36Gemini 3.1 Pro Previewno tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.5%
37Gemini 3.1 Pro Previewno tools · average of Standard (10 options) and Vision· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.5%
38GPT 5.2Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted · with tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.4%
39GPT 5.2Python · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.4%
40GPT 5.2with tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.4%
41Muse Spark· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim80.4%
42Kimi K2.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim
Values from 2 sources
80.1%
43Kimi K2.6Kimi K2.6 official benchmark table · thinking · temp 1.0 · top_p 1.0 · 262,144 context · with Python · max 50 steps · avg@3· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.80.1%
44GPT 5.2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim79.5%
45GPT 5.2no tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.5%
46GPT 5.2no tools · xhigh· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.5%
47GPT 5.2Opus 4.6 System Card capability summary · adaptive thinking · max effort · default sampling · 5 trials unless noted · no tools· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.5%
48GPT-5.6 Lunawith tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.5%
49Kimi K2.6Kimi K2.6 official benchmark table · thinking · temp 1.0 · top_p 1.0 · 262,144 context · no Python · official protocol · avg@3· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79.4%
50Qwen3.5 397B A17BQwen3.5 release/model card · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.79%

Showing 1–50 of 134 results.

Not ranked

The source marks these results as descriptive, not capability scores, so we list them by name instead of ranking them.

Model or systemScore
DeepSeek V4.1 Flash BaseEM · 4-shot· Reported by DeepSeek· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it.56.5%

Data credits: BenchBoard by SYuan03 (github.com/SYuan03/benchboard). MIT covers its code and data organization; no results licence is inferred. Numeric relays retain the original stating-source links. · Data by LLM Stats · Hugging Face official benchmark leaderboards; licence varies by benchmark