Chinese artificial intelligence company Zhipu AI said it aims to become a global leader in large language models by 2026, as it prepares to launch a new generation of its flagship model and step up investment in frontier research.

In an internal letter released on the first day of Zhipu’s Hong Kong stock market debut, founder and chief scientist Tang Jie said the company would soon roll out GLM-5, an upgraded version of its General Language Model series designed to handle more complex real-world tasks.

Tang said GLM-5 will be built through further scaling and the introduction of new technologies, enabling artificial intelligence systems to move beyond narrow functions and support longer and more sophisticated workflows.

He also said the widely used Transformer architecture, which has underpinned most large language models for nearly a decade, is showing limitations in areas such as long-context computation, memory and model updating, prompting the need to explore new model structures and scaling approaches.

Zhipu will establish a new internal research unit, X-Lab, to focus on cutting-edge exploration including new model architectures, cognitive paradigms and the incubation of new projects, Tang said. The lab will not be limited to software and may also pursue hardware-related research.

Tang added that the company plans to expand external investments, not only deepening cooperation with existing portfolio companies but also opening up new areas of strategic expansion.

Zhipu’s shares closed up 13.17% on their first day of trading in Hong Kong.