晚点独家丨智谱上市，唐杰内部信要求全面回归基础模型研究

Core Viewpoint - The article emphasizes that the next phase of the AI landscape will be determined by foundational aspects such as model architecture and learning paradigms, with 2026 expected to be a breakout year for AI replacing various jobs and tasks [3][16]. Group 1: Company Developments - On January 8, the company announced the upcoming release of its new model, GLM-5, as part of its ongoing pursuit of AGI (Artificial General Intelligence) [4]. - The company successfully completed its strategic goals for 2025, which included releasing models at different stages throughout the year, culminating in a top-ranking model by the end of the year [6][15]. - The company’s GLM-4.7 model was ranked first among domestic models and sixth globally, showcasing its competitive edge in the AI landscape [6]. Group 2: Strategic Focus Areas - The company plans to focus on three key technological directions for 2026: new model architecture design, more generalized reinforcement learning (RL) paradigms, and exploration of continuous learning and self-evolution for models [6][18]. - The company aims to enhance the capabilities of foundational models, integrating them with agent and domain-specific models, while also emphasizing the need for AI applications to return to first principles [7][16]. Group 3: Market Position and Future Outlook - The company has experienced significant growth, with its MaaS platform achieving an annual recurring revenue (ARR) of over 500 million, marking a 25-fold increase in just ten months [15]. - The emergence of competitors like DeepSeek has prompted the company to refocus on foundational model research, indicating a shift in strategy to maintain its market position [6][12]. - The company aspires to become a leading global player in the large model sector by 2026, with a commitment to making AI accessible and beneficial for all [16][20].