Deepseek-VL2

Deepseek-VL2

MoE Vision-Language, Now Easier to Access

585 followers

DeepSeek-VL2 are open-source vision-language models with strong multimodal understanding, powered by an efficient MoE architecture. Easily test them out with the new Hugging Face demo.

Deepseek-VL2 launches

Launch date
 DeepSeek-VL2
DeepSeek-VL2 MoE vision-language, now easier to access

Launched on February 10th, 2025