DeepSeek has released V4.1 Flash, rolling it out to web chat and mobile apps. The new model unifies the previous fast, expert, and vision modes into a single system: it automatically determines query complexity and activates vision when images are provided, eliminating the need for manual mode selection.

According to DeepSeek, V4.1 Flash outperforms V4 Pro in quality, cost, and speed. Until V4.1 Pro launches, it will handle all requests previously routed to V4 Pro.

Related: DeepSeek Begins Testing V4.1 Flash, DeepSeek V4 Pro with DSpark Released