OpenVLM MMMU - Humanities and Social Science: leaderboard
OpenCompass OpenVLM MMMU validation subcategory score for visual college-level questions in humanities and social science.
Metric: Accuracy (%). Source: huggingface.co. Saturation forecast: Estimated already saturated. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 85 | #152 |
| 2 | GPT-5 | 85 | #98 |
| 3 | GPT-4o ChatGPT | 82.5 | |
| 4 | GPT-4.1 (2025-04-14) | 81.7 | #250 |
| 5 | GPT-5 Mini (2025-08-07) | 80.8 | #172 |
| 6 | InternVL2.5-78B | 80.8 | #431 |
| 7 | InternVL3-78B | 80 | #352 |
| 8 | Qwen 2 VL 72B | 80 | #460 |
| 9 | Grok 2 (1212) | 80 | #496 |
| 10 | GPT-4.5 | 80 | #274 |
| 11 | InternVL3-38B | 79.2 | #405 |
| 12 | InternVL3-14B | 79.2 | #505 |
| 13 | Gemma 3 27B | 79.2 | #607 |
| 14 | Gemini 1.5 Flash (002) | 78.3 | #583 |
| 15 | InternVL3-8B | 77.5 | #620 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=openvlm-mmmu-humanities-and-social-science · How It Works · Data refreshed daily, snapshot 2026-10-20.