On February 11, China's domestic AI landscape witnessed significant developments, as Zhipu GLM-5, Minimax 2.5, and DeepSeek all rolled out new large-scale models on the same day. Among these, DeepSeek's update particularly drew attention. In this iteration, DeepSeek substantially boosted its context processing capacity, escalating it from the previous 128K to 1M—an almost eightfold surge. This advancement enables DeepSeek to process exceptionally long texts, such as the entire "Three-Body" trilogy, in a single pass, significantly broadening its range of potential applications. User evaluations have revealed marked enhancements in the new model's code generation, mathematical reasoning, and multi-turn conversational abilities. Although certain alterations in the interaction style have sparked discussions among users, the model's overall performance has garnered widespread acclaim.