DeepSeek: The official version of DeepSeek-V4-Pro will be released as soon as possible
22 hour ago / Read about 0 minute
Author:小编   

On July 31, 2026, DeepSeek released an update log announcing the official launch of the public beta for the DeepSeek-V4-Flash API. The model structure and size of this version remain consistent with the preview version, with only post-training reconfiguration. Meanwhile, the official stated that the official version of DeepSeek-V4-Pro will be released as soon as possible. DeepSeek-V4-Flash supports a context length of up to 1 million tokens and employs a hybrid attention architecture that combines compressed sparse attention with highly compressed attention to enhance long-context processing efficiency. With 284 billion parameters and an activation scale of 13 billion, it is a lightweight and efficient model.