DeepSeek unveils V4.1 flash intermediate version with enhanced multimodal capabilities

On September 8, 2026, DeepSeek announced the internal testing of its V4.1 Flash intermediate version, featuring a new model architecture, native multimodal support, improved performance, and reduced costs. Users can access the model by setting the model name to 'deepseek-v4.1-flash-expires-on-0910' without changing the base URL. The pricing remains the same as the 'deepseek-v4-flash' model, with a limit of 20 concurrent requests per account. The company also sought feedback on whether the V4.1 Flash intermediate version could fully replace the current DeepSeek V4 Pro. Earlier in July, the official DeepSeek-V4-Flash API was launched, significantly enhancing agent capabilities. On August 21, the DeepSeek-V4-Flash-Vision-Exp multimodal model was introduced, supporting image input in formats like JPEG, PNG, GIF, and WebP, enabling tasks such as image description, text recognition, and chart analysis.