MM STEM & Puzzle on MMMU (dev)
88.2ScoreQwen3-VL-32B
Evaluation results
| Method | Links | |
|---|---|---|
| Qwen3-VL-32BDecoding Mode=Thinking, Model Scale=32B2026.06 | 88.2 | |
| Opus-4.6Parameter Count=≈800B2026.06 | 83.3 | |
| K2.5Parameter Count=≈1T2026.06 | 78.7 | |
| Qwen3.5-PlusParameter Count=397B2026.06 | 77.3 | |
| GPT-5.4Parameter Count=≈1.5T2026.06 | 76.7 | |
| Qwen3-VL-8BDecoding Mode=Thinking, Model Scale=8B2026.06 | 74.2 | |
| Yuvion VL-32BDecoding Mode=Instruct, Model Scale=32B2026.06 | 74 | |
| Qwen3-VL-32BDecoding Mode=Instruct, Model Scale=32B2026.06 | 70.7 | |
| Yuvion VL-32BDecoding Mode=Reasoning, Model Scale=32B2026.06 | 70 | |
| Yuvion VL-8BDecoding Mode=Reasoning, Model Scale=8B2026.06 | 66 | |
| Qwen3-VL-8BDecoding Mode=Instruct, Model Scale=8B2026.06 | 63.3 | |
| Yuvion VL-8BDecoding Mode=Instruct, Model Scale=8B2026.06 | 60 |