Zhipu receives $5 billion investment to advance self-training glm model

On September 13, Zhipu, an AI company, announced securing approximately $5 billion in new financing. The funds will be allocated to develop the next-generation GLM foundation model, build a fully self-training system, and enhance computing infrastructure. Zhipu's R&D roadmap emphasizes 'fully self-training' as the core objective for the next-generation GLM, which will be trained within the environment created by the previous GLM version, enabling recursive self-improvement. The investment will support automated data generation and screening, task environment construction, and improved long-range reasoning capabilities. Additionally, engineering efforts will focus on domestic chip compatibility, operator development, and inference efficiency optimization. Zhipu's strategy addresses both model capability and computing efficiency, aiming to expand data and task spaces while maximizing output under existing hardware conditions. The $5 billion funding round marks a significant milestone in Zhipu's development of its next-generation model, establishing a comprehensive R&D foundation from algorithmic systems to engineering facilities. The company's dual focus on model capability and computing efficiency highlights its commitment to advancing efficient technological iteration paradigms. By integrating the next-generation GLM into a self-training cycle, Zhipu aims to create a sustainable and scalable development framework. This strategic investment underscores Zhipu's ambition to lead in AI innovation through advanced self-improvement mechanisms and optimized infrastructure.