On the evening of February 11, DeepSeek stealthily implemented updates across its web and app interfaces, commencing grayscale testing. This update notably extended the context length capacity from 128K Tokens to a staggering 1M Tokens, marking an almost eightfold enhancement. This advancement allows for the seamless processing of extensive texts, such as the entire 'The Three-Body Problem' series in a single instance. However, following the update, certain users observed abrupt shifts in the writing style and a deterioration in the interactive experience. Notably, DeepSeek ceased using user nicknames and modified its reply patterns, prompting a wave of collective dissatisfaction. Despite these concerns, DeepSeek maintains its competitive edge in terms of cost-efficiency, as its 1M text processing feature is priced at merely one-tenth of that of comparable rivals. At present, the updated version lacks support for visual comprehension or multimodal input, continuing to focus exclusively on textual and vocal interactions.