Key facts
- DeepSeek launched an experimental multimodal AI model, DeepSeek-V4-Flash-Vision-Exp.
- The model processes visual inputs and retains text capabilities.
- DeepSeek claims its multimodal agentic performance rivals Anthropic's Opus 4.8.
- The model is available to developers via an API.
- Anthropic is reportedly preparing for a potential IPO.
Chinese artificial intelligence startup DeepSeek has launched an experimental multimodal AI model, DeepSeek-V4-Flash-Vision-Exp, capable of processing visual data alongside text. This release marks an expansion beyond text-based models as the company competes with global rivals in developing general-purpose AI systems.
DeepSeek claims the new model retains the text-based capabilities of its V4 Flash predecessor, including reasoning and general knowledge, and that its multimodal agentic performance is comparable to Anthropic's Opus 4.8. The model is available to developers through an application programming interface (API).
The launch occurs amid intense competition in the AI sector. Reports indicate that Anthropic is preparing for a potential initial public offering (IPO), with investment banks such as Morgan Stanley, Goldman Sachs, and JPMorgan reportedly involved in the preparations. DeepSeek's existing DeepSeek-VL series focuses on visual-language capabilities, and the new model integrates these features into its V4 family.