DeepSeek releases V4.1 Flash beta version: Native multi-modal support, can replace the online Pro version

On September 8, 2026, DeepSeek released the beta version of V4.1 Flash. This version features a new model structure, native multi-modal support, and offers stronger capabilities, faster speed, and lower costs. Users can call the model by keeping the base_url unchanged and setting the model name to deepseek-v4.1-flash-expires-on-0910. Current billing rules are the same as those for deepseek-v4-flash, with a limit of 20 concurrent requests per account. DeepSeek officially asked in the feedback questionnaire whether this version could fully replace the online DeepSeek V4 Pro. Previously, the official API of DeepSeek-V4-Flash was launched for public testing at the end of July, and the Agent capabilities were significantly enhanced; on August 21, the DeepSeek-V4-Flash-Vision-Exp multi-modal visual understanding model was released, supporting input in JPEG, PNG, GIF, W…