(Yicai) Feb. 12 -- Shares of Knowledge Atlas Technology Joint Stock, better known as Zhipu AI, surged after the Chinese artificial intelligence developer released the latest version of its flagship GLM-5 large language model.

The shares [HKG: 2513] closed 29 percent higher at HKD402 (USD51.43) each in Hong Kong today, after earlier surging by as much as 42 percent. The stock has more than tripled in value since its first day of trading on Jan. 8.

GLM-5, which is open-source, prioritizes enhancing programming and agent capabilities, supports larger model scales and more complex reinforcement learning tasks, Beijing-based Zhipu noted. It has an asynchronous agent reinforcement learning algorithm, enabling the model to continuously learn from long-range interactions. 

The mysterious Pony Alpha model, which topped OpenRouter's model popularity rankings earlier this week, was in fact the latest iteration of GLM-5, Zhipu told Yicai.

For the first time, GLM-5 integrates DeepSeek Sparse Attention, a sparse attention mechanism, which maintains the quality of long text processing without loss while substantially lowering the model’s deployment cost, the company said.

AI model updates this year show them progressing from writing code and front-end tasks to handling engineering work and completing large projects. GLM-5's performance in practical programming tests is approaching that of Anthropic's Claude Opus 4.5, while it also excels in complex systems engineering and long-horizon agent tasks.

Furthermore, GLM-5 can be integrated into the viral open-source AI agent framework OpenClaw, providing users with an intelligent “intern” for various tasks, including website searches, organizing information, tweeting posts, and performs programming tasks.

Editor: Martin Kadiev