One model DeepSeek-VL, a state-of-the-art Vision-Language model designed for real-world applications, is now available. Supports multimodal tasks like logical diagrams, scientific literature, and more. Released for research and commercial use.