evals.report
BenchmarksLabsCompareRun guidesIn the wild
BenchmarksMultimodal

MMMU-Pro

The harder MMMU-Pro multimodal reasoning benchmark (college-level subject tasks with text and images); the variant current frontier models report.

MultimodalaccuracyHigher is better
ScoresAboutRun this benchmark

What is MMMU-Pro?

The harder MMMU-Pro multimodal reasoning benchmark (college-level subject tasks with text and images); the variant current frontier models report. evals.report tracks reported MMMU-Pro scores with the model, source, status, date, and run caveats attached — official leaderboard scores, vendor-reported launches, and clearly labeled community runs.

Top reported MMMU-Pro score: GPT-5.6 Sol — 83.0% (accuracy).

ModelLabScore↓Source modelStatusDate
GPT-5.6 SolOpenAI83.0%GPT-5.6 SolVerifiedJul 9, 2026Details
GPT-5.4OpenAI82.1%GPT-5.4 Thinking w/ toolsOfficialMar 5, 2026Details
Gemini 3 ProGoogle DeepMind81.0%Gemini 3.0 ProOfficialNov 18, 2025Details
GPT-5.6 TerraOpenAI80.7%GPT-5.6 TerraVerifiedJul 9, 2026Details
Gemini 3.1 Pro PreviewGoogle DeepMind80.5%Gemini 3.1 Pro Thinking (High)OfficialFeb 19, 2026Details
Muse SparkMeta80.4%Muse Spark ThinkingOfficialApr 8, 2026Details
GPT-5.2OpenAI80.4%GPT-5.2 Thinking w/o PythonOfficialDec 11, 2025Details
GPT-5.1OpenAI79.0%GPT-5.1 ThinkingOfficialNov 12, 2025Details
Qwen3.5-397B-A17BOpenAlibaba / Qwen79.0%Qwen3.5-397B-A17BUnverifiedFeb 16, 2026Details
GPT-5OpenAI78.4%GPT-5 w/ thinkingOfficialAug 7, 2025Details
GPT-5.6 LunaOpenAI78.4%GPT-5.6 LunaVerifiedJul 9, 2026Details
Doubao Seed 2.0 ProByteDance78.2%Seed2.0 Pro (0215)VerifiedFeb 14, 2026Details
MiMo-V2.5OpenXiaomi77.9 (MMMU-Pro)MiMo-V2.5VerifiedApr 22, 2026Details
Claude Opus 4.6Anthropic77.3%Claude Opus 4.6 w/ toolsOfficialFeb 5, 2026Details
o3OpenAI76.4%o3OfficialApr 16, 2025Details
Claude Sonnet 4.6Anthropic75.6%Claude Sonnet 4.6 w/ toolsOfficialFeb 17, 2026Details
Claude Opus 4.5Anthropic73.9%Claude Opus 4.5OfficialNov 24, 2025Details
InklingOpenThinking Machines Lab73.5%InklingVerifiedJul 15, 2026Details
Claude Sonnet 4.5Anthropic68.9%Claude Sonnet 4.5OfficialSep 29, 2025Details
Gemini 2.5 ProGoogle DeepMind68.0%Gemini 2.5 Pro 05-06OfficialMar 25, 2025Details
Amazon Nova 2 ProAmazon63.5%Nova 2 ProVerifiedDec 2, 2025Details
Amazon Nova 2 LiteAmazon61.8% (MMMU-Pro)Nova 2 LiteVerifiedDec 2, 2025Details
GPT-4oOpenAI51.9%GPT-4o (0513)OfficialMay 13, 2024Details
Claude 3.5 SonnetAnthropic51.5%Claude 3.5 SonnetOfficialJun 20, 2024Details

Each row reports the model’s accuracy on MMMU-Pro. Click a row for the full run context.