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
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 system | Score |
|---|---|---|
| 51 | Qwen3.7 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 79% |
| 52 | Qwen3.7 PlusQwen3.7-Plus release · published setting· Reported by Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 79% |
| 53 | Qwen3.6 Plus· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 78.8% |
| 54 | Qwen3.6 PlusQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.8% |
| 55 | Kimi K2.5· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 78.5% |
| 56 | Kimi K2.5official model-card base evaluation· Reported by Moonshot AI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.5% |
| 57 | GPT 5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 78.4% |
| 58 | GPT-5.6 Luna· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 78.4% |
| 59 | GPT-5.6 Lunano tools· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.4% |
| 60 | MiniMax M3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 78.1% |
| 61 | MiniMax M3official-aligned prompt· Reported by MiniMax Research· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78.1% |
| 62 | GPT-5.4 minixhigh · with Python· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 78% |
| 63 | MiMo V2.5· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 77.9% |
| 64 | MiMo V2.5Xiaomi MiMo-V2.5 official release/model card · published multimodal setting· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 77.9% |
| 65 | Claude Opus 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 77.3% |
| 66 | Claude Opus 4.6Opus 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. | 77.3% |
| 67 | Gemma 4 31B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.9% |
| 68 | Gemma 4 31B ITGemma 4 Technical Report · Table 6 · thinking · 1120 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 76.9% |
| 69 | Qwen3.5 122B-A10B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.9% |
| 70 | Gemini 3.1 Flash Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.8% |
| 71 | Gemini 3.1 Flash LiteGemini 3.1 Flash-Lite Model Card · high thinking · default sampling · single attempt/pass@1 · no tools· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 76.8% |
| 72 | MiMo V2 OmniXiaomi MiMo-V2-Omni release table · multimodal understanding and reasoning· Reported by Xiaomi MiMo· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 76.8% |
| 73 | GPT-5.4 mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.6% |
| 74 | GPT-5.4 minixhigh · without Python· Reported by OpenAI· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 76.6% |
| 75 | OpenAI o3· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76.4% |
| 76 | GPT-5.5 Instant· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 76% |
| 77 | Gemma 4 31B ITGemma 4 Technical Report · Table 12 · thinking · 280 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.8% |
| 78 | Hy4 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 75.8% |
| 79 | Qwen3.6 27B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 75.8% |
| 80 | Qwen3.6 27BQwen official release table · thinking mode unless benchmark requires otherwise· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.8% |
| 81 | Claude Sonnet 4.6· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 75.6% |
| 82 | Claude Sonnet 4.6image-cropping tool · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.6% |
| 83 | Qwen3.6 35B A3B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 75.3% |
| 84 | Qwen3.6 FlashQwen3.6-35B-A3B official model card · hosted API name Qwen3.6-Flash· Reported by Alibaba Qwen· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 75.3% |
| 85 | Claude Opus 4.7Gemini 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. | 75.2% |
| 86 | Qwen3.5 35B A3B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 75.1% |
| 87 | Qwen3.5 27B· Hugging Face Official Benchmarks data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 75% |
| 88 | Inkling 2 Mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 74.9% |
| 89 | Claude Sonnet 4.6Gemini 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. | 74.5% |
| 90 | Claude Sonnet 4.6no tools · adaptive thinking · max effort· Reported by Anthropic· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 74.5% |
| 91 | Inkling Small· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 74% |
| 92 | Muse Glimmer 30B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 74% |
| 93 | Muse Glimmer 30Bhigh reasoning· Reported by Meta· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 74% |
| 94 | Claude Opus 4.5Opus 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. | 73.9% |
| 95 | Claude Opus 4.6Opus 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. | 73.9% |
| 96 | Qwen3.5 Omni PlusQwen3.5-Omni release · published setting· Reported by Qwen, Qwen Team· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 73.9% |
| 97 | Gemma 4 26B A4B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 73.8% |
| 98 | Gemma 4 26B A4BGemma 4 Technical Report · Table 6 · thinking · 1120 vision tokens· Reported by Google DeepMind· via BenchBoard docs· Reported by the model's developer. The source relaying it did not verify it. | 73.8% |
| 99 | Inkling FP4· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 73.5% |
| 100 | Doubao Seed 1.8· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 73.2% |
Showing 51–100 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 system | Score |
|---|---|
| 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