DeepSeek officially announced that its API has been successfully updated to the same 1 million token context version as its App and web version, marking a significant technological advancement. Previously, DeepSeek's context capacity was only 128k tokens, but after this update, users can now enjoy longer context information, greatly enhancing the coherence and richness of conversations.

At the same time, DeepSeek's knowledge base has also been fully updated, now up to May 2025, which means users can accurately access the latest news up to April 2025 even without an internet connection. This advancement not only demonstrates DeepSeek's enhanced capabilities in information updates and processing, but also shows its commitment to providing better service to users.

image.png

It should be noted that DeepSeek is currently still a non-multimodal model, supporting only text and voice input, but not visual input. This limitation reminds users to choose the appropriate interaction method based on their needs during use.

DeepSeek's founder Liang Wenfeng previously revealed that the next generation flagship large model DeepSeek V4 will be released at the end of April. With continuous technological iteration, DeepSeek's features are also constantly improving, including the recently launched expert mode, which focuses on answering complex questions, further expanding its application scenarios.