On July 11, Tang Jie, founder of the large model company Zhipu, announced an internal letter stating that the company will launch the "Touch High (Mo Gao) Plan," strategically investing in four core engines - long-range tasks, autonomous intelligent agent systems, fully self-trained models, and extreme security governance - over the next two years to strive for the next generation of general artificial intelligence (AGI). Recently, Zhipu released the open-source model GLM-5.2, which supports a million (1M) context length and maintains leadership in long-range tasks, and is now open-sourced to all users under the MIT license.
In this strategic deployment, Zhipu clearly focuses its technological research and development on the planning and execution of long-range tasks, pushing models from instant Q&A towards grand projects. At the same time, by building thousands of intelligent agents with specialized skills, it accelerates the evolution of digital productivity toward the fully automated company (NPC) form. In terms of data resources, Zhipu will build a high-quality synthetic data factory, achieving fully self-training and code refactoring through AI game-based self-play. Furthermore, in response to the trend of super intelligence, the company will invest billions of resources to tackle "mechanical explainability," promoting the transition of black-box systems to transparent systems, while advancing super alignment and safety research in parallel.
This strategic move took place against the backdrop of significant fluctuations in Zhipu's market valuation. Since its listing on the Hong Kong Stock Exchange on January 8, 2026, Zhipu's stock price rose sharply within six months, reaching a high of HKD 2,410 on June 22, becoming the first large model company in China to exceed a market value of HKD 1 trillion. Subsequently, the market re-evaluated the narrative of large models, and as of the closing on July 10, its stock price had dropped to HKD 1,640, with the total market value shrinking to HKD 731.2 billion. Amid the industry's general acceleration in commercial monetization, Zhipu chose to "break through upwards" counterintuitively, highlighting the struggle and commitment of frontier AI companies within the dual constraints of technical physical limits and safety governance.
