General VQA on MMBench en (dev)
90.3AccuracyK2.5
Evaluation results
| Method | Links | |
|---|---|---|
| K2.5Parameter Count=≈1T2026.06 | 90.3 | |
| Qwen3.5-PlusParameter Count=397B2026.06 | 90.2 | |
| Qwen3-VL-32BDecoding Mode=Thinking, Model Scale=32B2026.06 | 89.5 | |
| Qwen3-VL-32BDecoding Mode=Instruct, Model Scale=32B2026.06 | 89 | |
| Qwen2.5-VL-72BParameters=72B2025.07 | 88.6 | |
| Opus-4.6Parameter Count=≈800B2026.06 | 87.8 | |
| Qwen3-VL-8BDecoding Mode=Thinking, Model Scale=8B2026.06 | 87.7 | |
| Yuvion VL-32BDecoding Mode=Instruct, Model Scale=32B2026.06 | 87.1 | |
| Yuvion VL-32BDecoding Mode=Reasoning, Model Scale=32B2026.06 | 86.6 | |
| Qwen3-VL-8BDecoding Mode=Instruct, Model Scale=8B2026.06 | 86.5 | |
| Yuvion VL-8BDecoding Mode=Reasoning, Model Scale=8B2026.06 | 85.6 | |
| TreeVGR-7BParameters=7B2025.07 | 84.4 | |
| Yuvion VL-8BDecoding Mode=Instruct, Model Scale=8B2026.06 | 84 | |
| GPT-5.4Parameter Count=≈1.5T2026.06 | 84 | |
| Qwen2.5-VL-7BParameters=7B2025.07 | 83.1 |