Massive Multi-discipline Multimodal Understanding Pro (MMMU-Pro)
Data verified
A harder multimodal benchmark for frontier models that combines text with images, diagrams, charts, and academic visual reasoning tasks.
Top models on MMMU-Pro — October 29, 2026
As of October 29, 2026, GPT-6.1 Terra leads the MMMU-Pro leaderboard with 85.5% , followed by Kimi K3.1 Pro (84.1%) and Gemini 3.1 Pro (83.9%).
1Closed
GPT-6.1 Terra
OpenAI
85.5%
Overall 80.46Context 1.05M
2Pending
Kimi K3.1 Pro
Moonshot AI
84.1%
Overall 74.61Context 1.05M
3Closed
Gemini 3.1 Pro
83.9%
Overall 64.41Context 1M
52 modelsMultimodal & Grounded40% of category scoreRefreshingUpdated October 29, 2026
Leaderboard (52 models)
52 of 52 rows
| Rank | Model / configuration | Score | Parameters (B) | Open / closed |
|---|---|---|---|---|
| 1 | GPT-6.1 TerraOpenAI | 85.5% | Not reported | Closed |
| 2 | Kimi K3.1 ProMoonshot AI | 84.1% | Not reported | Pending |
| 3 | Gemini 3.1 ProGoogle | 83.9% | Not reported | Closed |
| 4 | Gemini 3.5 FlashGoogle | 83.6% | Not reported | Closed |
| 5 | GPT-5.6 SolOpenAI | 83% | Not reported | Closed |
| 6 | Qwen3.8 MaxAlibaba | 82.3% | Not reported | Open |
| 7 | Kimi K3Moonshot AI | 81.6% | Not reported | Open |
| 8 | Seed 2.1 ProByteDance | 81.6% | Not reported | Closed |
| 9 | GPT-5.5OpenAI | 81.2% | Not reported | Closed |
| 10 | GPT-5.4OpenAI | 81.2% | Not reported | Closed |
| 11 | Gemini 3 ProGoogle | 81% | Not reported | Closed |
| 12 | GPT-5.6 TerraOpenAI | 80.7% | Not reported | Closed |
| 13 | Muse SparkMeta | 80.4% | Not reported | Closed |
| 14 | Seed 2.1 TurboByteDance | 80.1% | Not reported | Closed |
| 15 | GPT-5.2OpenAI | 79.5% | Not reported | Closed |
| 16 | Kimi K2.6Moonshot AI | 79.4% | Not reported | Open |
| 17 | dots3-note PreviewDots Studio | 79.1% | Not reported | Open |
| 18 | Qwen3.7 PlusAlibaba | 79% | Not reported | Closed |
| 19 | Qwen3.5 397BAlibaba | 79% | Not reported | Open |
| 20 | Qwen3.6 PlusAlibaba | 78.8% | Not reported | Closed |
| 21 | Kimi K2.5 (Reasoning)Moonshot AI | 78.5% | Not reported | Closed |
| 22 | Kimi K2.5Moonshot AI | 78.5% | Not reported | Open |
| 23 | GPT-5.6 LunaOpenAI | 78.4% | Not reported | Closed |
| 24 | MiniMax M3MiniMax | 78.1% | Not reported | Open |
| 25 | Grok 4.3xAI | 78.1% | Not reported | Closed |
| 26 | Pareto 26.9Unbiased | 78% | Not reported | Closed |
| 27 | MiMo-V2.5Xiaomi | 77.9% | Not reported | Closed |
| 28 | Claude Opus 4.6Anthropic | 77.3% | Not reported | Closed |
| 29 | Qwen3.5-122B-A10BAlibaba | 76.9% | Not reported | Open |
| 30 | Gemma 4 31BGoogle | 76.9% | Not reported | Open |
| 31 | GPT-5.4 miniOpenAI | 76.6% | Not reported | Closed |
| 32 | Step 5 PreviewStepFun | 76% | Not reported | Open |
| 33 | Qwen3.6-27BAlibaba | 75.8% | Not reported | Open |
| 34 | Qwen3.6-35B-A3BAlibaba | 75.3% | Not reported | Open |
| 35 | Grok 4.20xAI | 75.2% | Not reported | Closed |
| 36 | Qwen3.5-35B-A3BAlibaba | 75.1% | Not reported | Open |
| 37 | Qwen3.5-27BAlibaba | 75% | Not reported | Open |
| 38 | Inkling-SmallThinking Machines Lab | 74% | Not reported | Open |
| 39 | Muse Glimmer 30BMeta | 74% | Not reported | Open |
| 40 | Gemma 4 26B A4BGoogle | 73.8% | Not reported | Open |
| 41 | InklingThinking Machines Lab | 73.5% | Not reported | Open |
| 42 | Interfaze BetaInterfaze | 71.1% | Not reported | Closed |
| 43 | Claude Opus 4.5Anthropic | 70.6% | Not reported | Closed |
| 44 | Gemma 4 12BGoogle | 69.1% | Not reported | Open |
| 45 | GPT-5.4 nanoOpenAI | 66.1% | Not reported | Closed |
| 46 | Command A+Cohere | 63% | Not reported | Open |
| 47 | Gemma 4 E4BGoogle | 52.6% | Not reported | Open |
| 48 | Qwen2.5-VL-32BAlibaba | 49.5% | Not reported | Open |
| 49 | Mistral Small 3.1Mistral | 49.3% | Not reported | Open |
| 50 | LFM3-VL-2BLiquidAI | 45.7% | Not reported | Open |
| 51 | Gemma 4 E2BGoogle | 44.2% | Not reported | Open |
| 52 | LFM2.5-VL-3BLiquidAI | 30.5% | Not reported | Open |
AA-MMMU-Pro
BenchLM stores the Artificial Analysis MMMU-Pro result separately from the core MMMU-Pro lane because Artificial Analysis controls the evaluation configuration.
123 of 123 rows
| Source rank | Source model or configuration | Score | Source setup | Parameters (B) | Open / closed |
|---|---|---|---|---|---|
| 1 | Claude Opus 5.5Anthropic | 87.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 2 | GPT-6 AstraOpenAI | 86.9% | Artificial AnalysisOriginal result | Not reported | Closed |
| 3 | GPT-6.1 SolOpenAI | 86.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 4 | Gemini 3.8 FlashGoogle | 85.6% | Artificial AnalysisOriginal result | Not reported | Closed |
| 5 | Gemini 3.7 FlashGoogle | 85.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 6 | Claude Opus 5Anthropic | 84.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 7 | GPT-6.1 TerraOpenAI | 84.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 8 | Gemini 3.5 FlashGoogle | 84.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 9 | Kimi K3.1 ProMoonshot AI | 83.9% | Artificial AnalysisOriginal result | Not reported | Pending |
| 10 | GPT-5.6 SolOpenAI | 83.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 11 | Gemini 3.6 FlashGoogle | 83.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 12 | GPT-6 SolOpenAI | 82.9% | Artificial AnalysisOriginal result | Not reported | Closed |
| 13 | Qwen3.8 Max PreviewAlibaba | 82.8% | Artificial AnalysisOriginal result | Not reported | Closed |
| 14 | Gemini 3.1 ProGoogle | 82.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 15 | Muse Spark OpenMeta | 80.9% | Artificial AnalysisOriginal result | Not reported | Open |
| 16 | GPT-5.6 TerraOpenAI | 80.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 17 | Kimi K3Moonshot AI | 80.5% | Artificial AnalysisOriginal result | Not reported | Open |
| 18 | Qwen3.7 PlusAlibaba | 80.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 19 | Muse SparkMeta | 80.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 20 | Grok 4.5xAI | 80.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 21 | Gemini 3 ProGoogle | 80.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 22 | GPT-5.5OpenAI | 79.9% | Artificial AnalysisOriginal result | Not reported | Closed |
| 23 | Qwen3.8-Flash-NextAlibaba | 79.8% | Artificial AnalysisOriginal result | Not reported | Open |
| 24 | GPT-6 LunaOpenAI | 79.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 25 | Kimi K2.6Moonshot AI | 79.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 26 | Apodex 1.1Apodex | 79.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 27 | Apodex 1.1 MiniApodex | 79.2% | Artificial AnalysisOriginal result | Not reported | Open |
| 28 | MiniMax M3.1 FlashMiniMax | 79.1% | Artificial AnalysisOriginal result | Not reported | Closed |
| 29 | Ling 3.0 Flash VLInclusionAI | 79.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 30 | Gemini 3.5 Flash-LiteGoogle | 79.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 31 | Claude Opus 4.7 (Adaptive)Anthropic | 78.8% | Artificial AnalysisOriginal result | Not reported | Closed |
| 32 | GPT-5.6 LunaOpenAI | 78.6% | Artificial AnalysisOriginal result | Not reported | Closed |
| 33 | MiniMax M3MiniMax | 78.6% | Artificial AnalysisOriginal result | Not reported | Open |
| 34 | Gemini 3 FlashGoogle | 78.6% | Artificial AnalysisOriginal result | Not reported | Closed |
| 35 | GPT-5.3 CodexOpenAI | 78.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 36 | GPT-5.3-Codex-SparkOpenAI | 78.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 37 | GPT-5.4OpenAI | 78.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 38 | Grok 4.3xAI | 78.1% | Artificial AnalysisOriginal result | Not reported | Closed |
| 39 | Qwen3.6 PlusAlibaba | 78.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 40 | Claude Sonnet 5Anthropic | 77.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 41 | Qwen3.5 397BAlibaba | 77.3% | Artificial AnalysisOriginal result | Not reported | Open |
| 42 | Qwen3.5 397B (Reasoning)Alibaba | 77.3% | Artificial AnalysisOriginal result | Not reported | Open |
| 43 | DeepSeek V4.1 Flash 1017DeepSeek | 77.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 44 | Step 5 PreviewStepFun | 76.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 45 | Mistral Large 4Mistral | 76.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 46 | Claude Opus 4.7Anthropic | 76.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 47 | GPT-5.2-CodexOpenAI | 76.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 48 | Qwen3.8-27BAlibaba | 76.3% | Artificial AnalysisOriginal result | Not reported | Open |
| 49 | GPT-5.1OpenAI | 75.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 50 | Kimi K2.5 (Reasoning)Moonshot AI | 75.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 51 | Kimi K2.5Moonshot AI | 75.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 52 | Claude Opus 4.6 (Adaptive)Anthropic | 75.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 53 | Step 3.7 FlashStepFun | 75.3% | Artificial AnalysisOriginal result | Not reported | Open |
| 54 | Qwen3.5-122B-A10BAlibaba | 75.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 55 | Qwen3.5-27BAlibaba | 75.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 56 | Qwen3.6-35B-A3BAlibaba | 75.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 57 | Gemini 2.5 ProGoogle | 74.9% | Artificial AnalysisOriginal result | Not reported | Closed |
| 58 | Qwen3.6-27BAlibaba | 74.6% | Artificial AnalysisOriginal result | Not reported | Open |
| 59 | Muse Glimmer 30BMeta | 74.3% | Artificial AnalysisOriginal result | Not reported | Open |
| 60 | GPT-5 (medium)OpenAI | 74.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 61 | GPT-5 (high)OpenAI | 74.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 62 | Inkling-SmallThinking Machines Lab | 74.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 63 | Claude Opus 4.5 ThinkingAnthropic | 74.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 64 | InklingThinking Machines Lab | 73.5% | Artificial AnalysisOriginal result | Not reported | Open |
| 65 | Gemma 4 31BGoogle | 73.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 66 | GPT-5.4 miniOpenAI | 73.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 67 | MiMo-V2.6-FlashXiaomi | 73.1% | Artificial AnalysisOriginal result | Not reported | Open |
| 68 | GLM-5V-TurboZ.AI | 72.8% | Artificial AnalysisOriginal result | Not reported | Closed |
| 69 | Qwen3.5-35B-A3BAlibaba | 72.7% | Artificial AnalysisOriginal result | Not reported | Open |
| 70 | Claude Opus 4.6Anthropic | 72.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 71 | GPT-5.1-CodexOpenAI | 72.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 72 | GPT-5.1-Codex-MaxOpenAI | 72.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 73 | Claude Opus 4.5Anthropic | 71.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 74 | Claude Sonnet 4.6Anthropic | 70.6% | Artificial AnalysisOriginal result | Not reported | Closed |
| 75 | o3OpenAI | 70.1% | Artificial AnalysisOriginal result | Not reported | Closed |
| 76 | GPT-5 miniOpenAI | 70.1% | Artificial AnalysisOriginal result | Not reported | Closed |
| 77 | MiMo-V2-OmniXiaomi | 69.9% | Artificial AnalysisOriginal result | Not reported | Closed |
| 78 | Gemma 4 12BGoogle | 69.7% | Artificial AnalysisOriginal result | Not reported | Open |
| 79 | Gemma 4 26B A4BGoogle | 69.2% | Artificial AnalysisOriginal result | Not reported | Open |
| 80 | o4-miniOpenAI | 69.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 81 | Grok 4xAI | 68.8% | Artificial AnalysisOriginal result | Not reported | Closed |
| 82 | Claude 4.1 Opus ThinkingAnthropic | 67.9% | Artificial AnalysisOriginal result | Not reported | Closed |
| 83 | Gemini 2.5 FlashGoogle | 65.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 84 | GPT-5.4 nanoOpenAI | 65.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 85 | Qwen3.5-4BAlibaba | 65.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 86 | Mistral Medium 3.5 128BMistral | 64.9% | Artificial AnalysisOriginal result | Not reported | Open |
| 87 | Grok 4.1 Fast (Reasoning)xAI | 63.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 88 | Command A+Cohere | 63.2% | Artificial AnalysisOriginal result | Not reported | Open |
| 89 | Claude 4 SonnetAnthropic | 62.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 90 | Llama 4 MaverickMeta | 62.1% | Artificial AnalysisOriginal result | Not reported | Open |
| 91 | Grok 4 Fast (Reasoning)xAI | 61.8% | Artificial AnalysisOriginal result | Not reported | Closed |
| 92 | GPT-4.1OpenAI | 61.2% | Artificial AnalysisOriginal result | Not reported | Closed |
| 93 | GPT-5 nanoOpenAI | 61.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 94 | Qwen3-Omni-30B-A3B-ThinkingAlibaba | 60.2% | Artificial AnalysisOriginal result | Not reported | Open |
| 95 | Claude 3.7 SonnetAnthropic | 60.1% | Artificial AnalysisOriginal result | Not reported | Closed |
| 96 | GPT-4.1 miniOpenAI | 58.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 97 | Mistral Small 4Mistral | 56.8% | Artificial AnalysisOriginal result | Not reported | Open |
| 98 | Mistral Large 3Mistral | 55.7% | Artificial AnalysisOriginal result | Not reported | Closed |
| 99 | Qwen3-Omni-30B-A3B-InstructAlibaba | 55.5% | Artificial AnalysisOriginal result | Not reported | Open |
| 100 | Gemini 1.5 ProGoogle | 55.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 101 | Gemini 2.5 Flash-LiteGoogle | 54.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 102 | Nemotron 3 Nano Omni 30B A3BNVIDIA | 53.2% | Artificial AnalysisOriginal result | Not reported | Open |
| 103 | Mistral Medium 3Mistral | 53.0% | Artificial AnalysisOriginal result | Not reported | Closed |
| 104 | Llama 4 ScoutMeta | 52.9% | Artificial AnalysisOriginal result | Not reported | Open |
| 105 | Gemma 4 E4BGoogle | 51.4% | Artificial AnalysisOriginal result | Not reported | Open |
| 106 | Ministral 3 14B (Reasoning)Mistral | 49.8% | Artificial AnalysisOriginal result | Not reported | Open |
| 107 | Ministral 3 14BMistral | 49.8% | Artificial AnalysisOriginal result | Not reported | Open |
| 108 | Grok 4.1 FastxAI | 48.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 109 | Gemini 1.5 FlashGoogle | 48.4% | Artificial AnalysisOriginal result | Not reported | Closed |
| 110 | Gemma 3 27BGoogle | 48.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 111 | Mistral Small 4 (Reasoning)Mistral | 46.5% | Artificial AnalysisOriginal result | Not reported | Open |
| 112 | Ministral 3 8B (Reasoning)Mistral | 46.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 113 | Ministral 3 8BMistral | 46.0% | Artificial AnalysisOriginal result | Not reported | Open |
| 114 | Claude 3.5 HaikuAnthropic | 45.6% | Artificial AnalysisOriginal result | Not reported | Closed |
| 115 | Gemma 4 E2BGoogle | 44.6% | Artificial AnalysisOriginal result | Not reported | Open |
| 116 | Nova ProAmazon | 44.3% | Artificial AnalysisOriginal result | Not reported | Closed |
| 117 | GPT-4o miniOpenAI | 41.5% | Artificial AnalysisOriginal result | Not reported | Closed |
| 118 | GPT-4.1 nanoOpenAI | 40.1% | Artificial AnalysisOriginal result | Not reported | Closed |
| 119 | Ministral 3 3B (Reasoning)Mistral | 38.1% | Artificial AnalysisOriginal result | Not reported | Open |
| 120 | Ministral 3 3BMistral | 38.1% | Artificial AnalysisOriginal result | Not reported | Open |
| 121 | Claude 3 HaikuAnthropic | 30.8% | Artificial AnalysisOriginal result | Not reported | Closed |
| 122 | LFM2.5-VL-1.6B-ExtractLiquidAI | 26.5% | Artificial AnalysisOriginal result | Not reported | Open |
| 123 | Phi-4 Multimodal InstructMicrosoft | 14.5% | Artificial AnalysisOriginal result | Not reported | Open |
According to BenchLM.ai, GPT-6.1 Terra leads the MMMU-Pro benchmark with a score of 85.5%, followed by Kimi K3.1 Pro (84.1%) and Gemini 3.1 Pro (83.9%). The top models are clustered within 1.6 points, suggesting this benchmark is nearing saturation for frontier models.
52 models have been evaluated on MMMU-Pro. The benchmark falls in the Multimodal & Grounded category. The Multimodal & Grounded leaderboard ranks models by a weighted category score, and MMMU-Pro contributes 40% of it. It does not enter the overall BenchAlign v5.8 ranking.
About MMMU-Pro
Year
2024
Tasks
Multimodal academic reasoning
Format
Image + text question answering
Difficulty
Frontier multimodal
MMMU-Pro extends the original MMMU setup with more difficult multimodal questions and stronger separation at the top end of the model market.
Freshness and provenance
Version
MMMU-Pro 2024
Refresh cadence
Annual
Staleness state
Refreshing
Question availability
Public benchmark set
Refreshing
BenchLM uses freshness metadata to decide whether a benchmark should still be treated as a strong differentiator, a benchmark to watch, or a display-only reference. For the full scoring policy, see the BenchLM methodology page.
Questions
What does MMMU-Pro measure?
A harder multimodal benchmark for frontier models that combines text with images, diagrams, charts, and academic visual reasoning tasks.
Which model scores highest on MMMU-Pro?
GPT-6.1 Terra by OpenAI currently leads with a score of 85.5% on MMMU-Pro.
How many models are evaluated on MMMU-Pro?
52 AI models have published results on MMMU-Pro in the BenchLM catalog.
Last updated: October 29, 2026 · BenchLM version MMMU-Pro 2024
Know when it’s worth switching models
The model to choose, the cheaper alternative, and the release we would wait on.
Read a sample issueJoin 6,100+ subscribers.
One email each week. Unsubscribe anytime.