Qwen

/

Qwen3-VL-32B-Thinking-FP8

like 20

Follow

Qwen 63.8k

I compared scores across many models:

RED color - high score
BLUE color - significantly low score
GREEN background - MOE

My conclusions:

• Dense models has a bit higher scores than MOE, but much slower
• Overall, it looks like QWEN3-VL models are very competitive to closed state-of-the-are models

I think we need more coding scores.