Seek .(US:SKLTY) Feng Huang Wang·2026-02-11 10:37Core Insights - DeepSeek has initiated a key update with a significant enhancement in its model architecture, moving from a context window of 128K to 1M tokens, which allows for processing longer texts comparable to international products like GPT-5 and Gemini3Pro [1] - The model's knowledge base has been updated to include information up to May 2025, and it can accurately output news events as far ahead as April 2025 [1] - User feedback indicates that the new model exhibits a more "enthusiastic and nuanced" language style, enhancing the user interaction experience [1] Group 1 - DeepSeek has begun gray testing for its updated model on both web and app platforms [1] - The new model's context window allows it to handle the entire "Three-Body" trilogy in a single processing instance [1] - The upgrade does not include multimodal visual understanding capabilities, focusing instead on text and voice interactions [1] Group 2 - DeepSeek has been actively hiring for multiple core technical positions, including deep learning researchers and engineers, indicating a focus on advancing its large language model (LLM) capabilities [2] - The company is open to various recruitment channels, including campus recruitment and internships, to fill these positions [2] - There is speculation that the current version being tested may correspond to the previously rumored "DeepSeek V4" or an enhanced version of V3.2 [2]