MMVet: leaderboard
218 open-ended questions over 200 images on recognition, OCR, knowledge, language generation, spatial awareness and math in 16 combinations, from NUS and Microsoft (2023); GPT-4 scores each answer.
Metric: Score. Source: arxiv.org. Source: huggingface.co. Source: rank.opencompass.org.cn. 284 models tracked.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Gemini 2.5 Pro | 83.3 | #152 |
| 2 | InternVL3-8B | 82.8 | #620 |
| 3 | InternVL3-38B | 81.1 | #405 |
| 4 | InternVL3-78B | 80.7 | #352 |
| 5 | InternVL3-14B | 80.5 | #505 |
| 6 | Doubao-1.5-Pro | 79.8 | #287 |
| 7 | GPT-4.1 (2025-04-14) | 78.8 | #250 |
| 8 | GPT-5 | 77.6 | #98 |
| 9 | GPT-4o ChatGPT | 76.9 | |
| 10 | GPT-4.5 | 75.3 | #274 |
| 11 | Gemini 1.5 Pro (002) | 74.6 | #349 |
| 12 | GPT-5 Mini (2025-08-07) | 74.6 | #172 |
| 13 | GPT-4.1 Mini | 74.3 | #355 |
| 14 | Qwen 2 VL 72B | 73.9 | #460 |
| 15 | Gemini 2.0 Flash | 73.6 | #342 |
No result here: #3 Claude Opus 5.5, #6 GPT-6 Astra, #7 GPT-6.1 Sol.
Interactive version: theaggregate.ai/benchmark?slug=mmvet · How It Works · Data refreshed daily, snapshot 2026-10-20.