MMBench (Chinese, test)
Multimodal · 2023-07-12
The Chinese-language version of MMBench's v1.0 test split: multiple-choice multimodal questions over 20 abilities, scored with CircularEval.
Top models (higher is better)
| Model | Score |
|---|---|
| InternVL2.5-78B | 88.5 |
| Qwen2.5-VL-72B-Instruct | 87.9 |
| Qwen2.5-VL-7B-Instruct | 83.4 |
| GPT-4o (2024-05-13) | 82.1 |
| Qwen2 VL 7B Instruct | 80.5 |
| Qwen2.5-VL-3B-Instruct | 78.1 |
| Qwen2 VL 2B Instruct | 73.5 |