The global downloads of domestic large models have exceeded 10 billion times, and trillion-parameter open-source models are emerging frequently. China's artificial intelligence is experiencing a comprehensive breakthrough across the entire industry chain.

Jiang Yi, director of the Policy Research Office and spokesperson of the National Development and Reform Commission, stated at a press conference on July 31 that in the first half of this year, China's artificial intelligence has made significant progress in independent innovation. Domestic companies such as DeepSeek and Moonshot AI have successively released open-source large models with parameter scales reaching "trillion-level."

Jiang Yi emphasized that in the next step, efforts will be made to balance high-quality development with high-level security, ensuring that artificial intelligence runs fast and steadily. Specifically, efforts will focus on three aspects: first, accelerating independent innovation, focusing on frontier directions such as fundamental theories of large models, training and inference efficiency, multimodal capabilities, and intelligent agents; second, emphasizing application orientation, accelerating the layout and construction of national artificial intelligence application pilot bases; third, deepening the ecosystem, vigorously cultivating intelligent derivative enterprises, and accelerating the legislative process of the Artificial Intelligence Law, striving to contribute a governance model with Chinese characteristics.

In addition, the official also pointed out that efforts will be strengthened to monitor and prevent risks in artificial intelligence, improve technical monitoring, risk early warning, and emergency response systems, and adhere to the principle that technology is led by people, used by people, and for the benefit of people, properly addressing the impact of artificial intelligence on areas such as economic operations and employment distribution.