Recently, "agents" and "tokens" have gained popularity, especially in open-source AI projects, where they are giving rise to new business models and market opportunities. As this technological trend continues, China's large model technology has made exponential progress and its applications are rapidly expanding. This is refreshing the intelligent economy and gradually becoming the cornerstone of global innovation.

As a driving force behind the current technological revolution and industrial transformation, AI stands at a crucial development juncture. For example, in terms of programming capabilities, AI demonstrates significant potential for application in code design, writing, testing, and deployment. In some knowledge-based tasks, it can even reach the level of human experts, prompting industry predictions that AI will further boost societal productivity in the foreseeable future and become a reliable assistant in people's work and life.

Yang Zhilin, founder and CEO of Moonshot AI, said that China's Internet and digital economy have a huge number of users and diverse application scenarios. These advantages have continued into the AI era. In March this year, the average daily token calls in China exceeded 140 trillion, growing more than 1,000-fold in two years, while the number of generative AI users exceeded 600 million, with a penetration rate of 42.8 percent.

More importantly, China has already produced a number of self-developed large models that can rival the world's best, becoming an indispensable variable in AI innovation. The rise of Chinese large models is reflected not only in quantity, but also in multiple aspects such as quality, application depth and global influence.

In 2025, China's open-source models topped the global download charts. The new technology not only enables a wider range of domestic chips to be effectively used for large model inference, but also reduces the overall operating cost of large models.

A large number of Chinese open-source models are becoming the "benchmark" in global model evaluation and an important sample for overseas research institutions.

The homegrown large model Kimi, for example, has gained a foothold in the global market due to its high cost-effectiveness. It has become a "productivity tool" for many overseas developers and users, with a significant increase in paid usage volume and other metrics. Chinese open-source models can be widely applied and subsequently form an ecosystem.