DeepSeek Merges Three Modes into One Smart Mode, V4.1 Flash Coming Soon

Chinese AI company DeepSeek announced it is merging its Fast, Expert, and Vision modes into one unified Smart Mode, so users no longer need to switch manually. In the new mode, the model can automatically detect query complexity, activate vision capabilities when image input is detected, and adjust processing power based on task difficulty.

Previously, the multimodal vision model DeepSeek-V4-Flash-Vision-Exp launched on the DeepSeek API platform, supporting three call formats. DeepSeek plans to officially release the V4.1 Flash model around September 10, 2026 Beijing time.

According to tests, V4.1 Flash has surpassed V4 Pro across all metrics including performance, cost, speed, and total time. Its internal beta version uses a new model architecture with native multimodal support, offering stronger capabilities, faster speed, and lower cost.