Core Insights - The core focus of the company is on advancing model architecture and learning paradigms, which are seen as critical for the next phase of AI development. The year 2026 is anticipated to be a breakthrough year for AI in replacing various jobs and tasks [1][19]. Group 1: Company Developments - The company announced the upcoming release of its new model, GLM-5, on January 8, coinciding with its public listing. The founder emphasized the importance of creating useful theories, technologies, or products in the pursuit of AGI [1][19]. - The company successfully launched its foundational model GLM-4.7 on December 23, which ranked first among domestic models and tied for sixth globally with Claude 4.5 Sonnet [2][20]. - The company achieved significant revenue growth, with its MaaS platform's annual recurring revenue (ARR) increasing from 20 million to 500 million (a 25-fold increase) within ten months following the GLM-4.7 release [29]. Group 2: Strategic Focus and Goals - The company plans to focus on three key technical directions for 2026: new model architecture design, more generalized reinforcement learning (RL) paradigms, and exploration of continuous learning and self-evolution [2][20]. - The company aims to become a leading global player in large models by 2026, emphasizing the need for innovative approaches and a commitment to long-term goals rather than short-term gains [31][32]. - The establishment of a new department, X-Lab, is intended to foster innovative exploration and collaboration among young talents, focusing on groundbreaking ideas that align with the pursuit of AGI [32][33]. Group 3: Industry Context - The emergence of DeepSeek has posed a challenge to the company, prompting a strategic return to foundational model research [2][20]. - The company recognizes the need for a paradigm shift in AI applications, suggesting that the future of AI will not necessarily require new applications but will return to fundamental principles [2][20]. - The company acknowledges the rapid evolution of AI since the release of ChatGPT, indicating a lack of consensus in the industry and a need for a more structured approach moving forward [2][20].
智谱上市，唐杰内部信要求全面回归基础模型研究