Multi-task Multimodal Understanding on MMT-Bench (val)
72.7ScoreGPT-5
Evaluation results
| Method | Links | |
|---|---|---|
| GPT-5Inference mode=Thinking mode2026.04 | 72.7 | |
| Qwen3-OmniInference mode=Thinking mode, Size=30B-A3B2026.04 | 70.9 | |
| Gemini 2.5 FlashInference mode=Thinking mode2026.04 | 70.7 | |
| Qwen3-OmniSize=30B-A3B, mode=instruct2026.04 | 70.4 | |
| Gemini 2.5 FlashSize=-, mode=instruct2026.04 | 70 | |
| MiniCPM-o 4.5Inference mode=Thinking mode, Size=9B2026.04 | 69.7 | |
| MiniCPM-o 4.5Size=9B, mode=instruct2026.04 | 69.7 | |
| Qwen3-VLInference mode=Thinking mode, Size=8B2026.04 | 68.1 | |
| InternVL3.5Size=8B, mode=instruct2026.04 | 66.7 | |
| Qwen3-VLSize=8B, mode=instruct2026.04 | 60.9 |