On September 8, DeepSeek announced in its official communication group that the intermediate version of V4.1 Flash has entered beta testing. This model adopts a new architecture and natively supports multimodality, offering enhanced capabilities, improved speed, and reduced costs. Developers can call the model by keeping the base_url unchanged and setting the model name to deepseek-v4.1-flash-expires-on-0910. The billing method remains the same as V4 Flash, with a limit of 20 concurrent requests per account. The beta test expires on September 10 and is aimed at functional verification and performance testing, rather than handling large-scale operations. Initial feedback indicates a significant speed improvement in the new model, with SVG code generation being 6 times faster and long-context retrieval over 5 times faster. The official statement mentions lower costs, but developers report that this is not directly reflected, possibly due to increased token consumption resulting from the faster speed. The advantages of improved underlying inference efficiency or reduced computational costs require more testing to verify if they can translate into reduced billing.