Qwen3 VL 32B Instruct
Alibaba · 2025-10-21 · 33.4B parameters
Qwen3-VL-32B-Instruct is the instruction-tuned, dense 32B vision-language model in Alibaba's Qwen3-VL family. It shares the family's image/video understanding, multilingual OCR, spatial grounding, and visual-agent capabilities while using all parameters per forward pass. Its context is 262,144 tokens natively and expandable to 1 million tokens.