It is reported that DeepSeek will release a V4 multimodal model next week, which natively supports AI generation of images, videos, and texts

On February 28th, the Financial Times (FT) published a blog post reporting that DeepSeek will release a new multimodal big language model V4 next week, which natively supports image, video, and text generation functions. This is the first major version update of DeepSeek since the launch of the R1 inference model in January 2025, which is expected to further fill the market demand for low-cost, open-source models in China.