Benchmark
MMMU
Vision benchmarks measure how well models understand images. Scores are percentages. Higher is better.
- Models measured
- 76
- Scores recorded
- 84
- Sources
- 4
- IQ weight
- 1.31
Score distribution
- 66.7%
- Mean
- 69.4%
- Median
- 14.3%
- Std dev
- 50.7%
- 10th percentile
- 82.9%
- 90th percentile
- 0%
- Min
This benchmark has a weight of 1.31 for chat models, measured over 71 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.
Leaderboard
| # | Model | Score |
|---|---|---|
| 51 | Grok 2 mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 63.2% |
| 52 | Mistral Small 3.2 24B Instruct2506· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 62.5% |
| 53 | Gemini 1.5 Flash· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 62.3% |
| 54 | GPT-4 Turbo· WildVision Arena data | 62.3% |
| 55 | gemini-1.5-pro-latest· WildVision Arena data | 62.2% |
| 56 | Nova Pro· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 61.7% |
| 57 | Claude 3 Opus· WildVision Arena data | 59.4% |
| 58 | GPT-4o mini· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claimValues from 2 sources
| 59.4% |
| 59 | Mistral Small 3.1 24b Instruct 2503· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 59.3% |
| 60 | Qwen2.5-Omni 7B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 59.2% |
| 61 | Qwen2.5-VL 7B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 58.6% |
| 62 | GPT-4-vision-preview· WildVision Arena data | 56.8% |
| 63 | LLaVA-OneVision Qwen2 72B OV· WildVision Arena data | 56.8% |
| 64 | Reka Core· WildVision Arena data | 56.3% |
| 65 | Nova Lite· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 56.2% |
| 66 | Gemini 1.5 Flashlatest· WildVision Arena data | 56.1% |
| 67 | GPT-4.1 nano· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 55.4% |
| 68 | Phi 4 Multimodal Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 55.1% |
| 69 | Qwen2 VL 7b Instruct· WildVision Arena data | 54.1% |
| 70 | Gemini 1.5 Flash-8B· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 53.7% |
| 71 | Reka Flash· WildVision Arena data | 53.3% |
| 72 | Claude 3 Sonnet· WildVision Arena data | 53.1% |
| 73 | Pixtral 12B2409· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 52.5% |
| 74 | Qwen-VL Max· WildVision Arena data | 51.4% |
| 75 | DeepSeek VL2· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 51.1% |
| 76 | LLaVA 1.6 34B· WildVision Arena dataValues from 2 sources
| 51.1% |
| 77 | Llama 3.2 11B Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 50.7% |
| 78 | Claude 3 Haiku· WildVision Arena data | 50.2% |
| 79 | Gemini Pro Vision· WildVision Arena data | 47.9% |
| 80 | gemini-1.0-pro· LLM Stats data | 47.9% |
| 81 | Qwen-VL Plus· WildVision Arena data | 45.2% |
| 82 | Phi-3.5 Vision Instruct· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim | 43% |
| 83 | Phi-3 Vision 128K Instruct· WildVision Arena data | 40.4% |
| 84 | GPT-3.5 Turbo· LLM Stats data | 0% |
Showing 51–84 of 84 results.
Data credits: WildVision Arena, MIT (licence) · 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