DeepSeek变冷淡了

Core Viewpoint - DeepSeek has significantly upgraded its flagship model, increasing the context window from 128K Tokens to 1M Tokens, allowing for a substantial enhancement in processing capacity and information retention during interactions [5][6]. Group 1: Model Upgrade Details - The new 1M Tokens context window enables DeepSeek to process approximately 750,000 to 900,000 English letters or around 80,000 to 150,000 lines of code in a single interaction [6]. - This upgrade allows DeepSeek to read and understand the entire "Three-Body Problem" trilogy (approximately 900,000 words) and perform macro analysis or detailed retrieval within minutes [6]. - The knowledge base of DeepSeek is set to be updated from mid-2024 to May 2025, although the current version does not support visual understanding or multimodal input, focusing solely on text and voice interactions [6]. Group 2: User Feedback and Reactions - Users have reported changes in the model's writing style post-update, describing it as more formal and less personal, leading to dissatisfaction among some who feel it has lost its previous empathetic touch [7][8]. - There is a growing call among users for DeepSeek to maintain its depth of thought and emotional understanding while enhancing technical capabilities, with some reverting to older versions of the application [8]. - The current gray version is not officially labeled as "DeepSeek-V4" and is perceived as a test version that prioritizes speed over quality, preparing for the anticipated V4 release in February 2026 [9].