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 made waves with their R1 language model, but their multimodal capabilities (especially image understanding) are not good enough:
But, they are rapidly evolving. DeepSeek-VL2, their new open-source family of Mixture-of-Experts (MoE) vision-language models, is a big step forward, achieving strong performance with a much smaller activated parameter count, thanks to its MoE design.
And the exciting news: there is a new Hugging Face Spaces demo – you can now try these models without needing to deploy heavily (normally you would need more than 80GB of GPU resources, which is almost impossible for most of us)
So check it out and see what DeepSeek brings next to suprise everyone :)
Great thanks for all of the developers in Deepseek. Deepseek R1 REALLY saves a lot of time on research tasks. And I would definitely recommend it to my colleges.
Although the server is mostly busy, the outcomes have consistently been great. I won't go as far as to say its as great as Chat GPT, but DeepSeek has a lot of potential.
DeepSeek-VL2 looks like an exciting advancement in the realm of vision-language models! Its strong multimodal understanding, powered by the efficient MoE architecture, is impressive and opens up many possibilities for applications in various fields, such as AI-driven content creation and interactive systems. The accessibility through the Hugging Face demo makes it easy for developers and researchers to experiment with and leverage its capabilities. Congrats on achieving the #2 ranking for the day! Looking forward to seeing how this model evolves and impacts the community!
Deepseek's greatest value is that R1 cost nothing compared to Chatgpt o1; There are still no good use cases for "middle detail" vision models yet (Good enough to identify things in pictures but not good enough to drive a car etc.)
Awesome release though.
DeepSeek-VL2 seems like a powerful tool for multimodal understanding, and the fact that it's open-source makes it even more accessible for developers and researchers!
I’m excited to see how it will transform industries with its powerful multimodal capabilities. Its efficiency and versatility open up many possibilities for practical applications. I look forward to witnessing its impact and the innovative solutions it will bring in the future.
I recommend DeepSeek-VL2! These are open models with powerful multimodal understanding that make interacting with images and text incredibly convenient and efficient. The models are based on an efficient MoE architecture, which significantly improves performance.
Always good to see more updates from DeepSeek! Let the competition begin, so everyone will be eager to come out with more interesting stuff more frequently!
Awesome, that's good news, I did not like the "No text extracted" error when i tried to upload images of any kind...
But, they are rapidly evolving. DeepSeek-VL2, their new open-source family of Mixture-of-Experts (MoE) vision-language models, is a big step forward, achieving strong performance with a much smaller activated parameter count, thanks to its MoE design. And the exciting news: there is a new Hugging Face Spaces demo – you can now try these models without needing to deploy heavily (normally you would need more than 80GB of GPU resources, which is almost impossible for most of us) So check it out and see what DeepSeek brings next to suprise everyone :)