DeepSeek has begun a limited-time beta of V4.1 Flash, an interim model that uses a new architecture and natively supports multimodal capabilities. DeepSeek said the model offers stronger performance and faster generation at a lower cost, but has not presented it as a formal release.
Developers can access the model through DeepSeek’s existing API by selecting deepseek-v4.1-flash-expires-on-0910. Beta pricing is the same as for V4 Flash, each account supports up to 20 concurrent requests, and the model is scheduled to go offline on Sept. 10. [APPSO, in Chinese]