#63 · Primary category： Foundation Models
DeepSeek-VL2
DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding
Project last updated：02/26/25
GitHub Stars
5.4K
Forks
1.8K
Contributors
10
License
MIT
Why we included this project
DeepSeek-VL2 is an open family of vision-language models built around a mixture-of-experts design, so the 1.0B, 2.8B, and 4.5B variants only activate a modest share of their parameters on each request. That keeps the compute bill closer to a small model while still covering visual question answering, OCR, and object localization, which matters for teams that want multimodal capability without paying for a dense frontier model. The repo has a straightforward Hugging Face integration and a short Python path: load a checkpoint, pass it one or several images, and get back text plus bounding boxes for grounding tasks. It is a practical pick for document understanding, visual search, or image captioning when you want a model you can run and fine-tune yourself.
Articles for this project
No articles for this project yet.
To suggest a topic or contribute an article, contact us.
Related projects in this category
transformers
🤗 Transformers: the model-definition framework for state-of-the-art machine learning models in text, vision, audio, and multimodal models, for both inference and training.
CLIP
CLIP (Contrastive Language-Image Pretraining), Predict the most relevant text snippet given an image
MiniCPM-V
A Pocket-Sized MLLM for Ultra-Efficient Image and Video Understanding on Your Phone
generative-models
Generative Models by Stability AI
unilm
Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities