DeepSeek Begins Testing V4.1 Flash

DeepSeek has launched a test phase for DeepSeek-V4.1-Flash, an intermediate model featuring a new architecture and native multimodal support. The company claims improved answer quality, faster inference, and lower costs compared to the previous version. To access the API, users keep the same base_url and specify the model as deepseek-v4.1-flash-expires-on-0910.

During the testing period, pricing matches the existing deepseek-v4-flash rates, with a limit of 20 concurrent requests per account.

Announcement