MMBench v1.1 (English, test)
Multimodal
MMBench v1.1 is the revised MMBench release with cleaned questions; this row is its English test split, scored with CircularEval.
Top models (higher is better)
| Model | Score |
|---|---|
| Qwen2.5-VL-72B-Instruct | 88.4 |
| InternVL2.5-78B | 87.4 |
| GPT-4o (2024-05-13) | 83.1 |
| Qwen2.5-VL-7B-Instruct | 82.6 |
| Qwen2 VL 7B Instruct | 80.7 |
| MiniCPM-o 2.6 | 78.0 |
| MiniCPM-V 2.6 | 78.0 |
| GPT-4o Mini | 76.0 |
| Qwen2 VL 2B Instruct | 72.2 |