Reference News Network, December 31 report (by Zhang Manzi) In the latest "AA Intelligence Index" list released by the global well-known large model evaluation institution "Artificial Intelligence Analysis Network" on December 28, the self-developed Chinese GLM-4.7 model ranked first among open-source models with a comprehensive score of 68 points.
Over the past year, the global artificial intelligence (AI) field has seen a "China fever". Since DeepSeek became popular at the beginning of 2025, China's large models have been rapidly entering the global top tier, thanks to breakthroughs in technological innovation, application implementation, and ecological building across multiple dimensions. A joint report by the Massachusetts Institute of Technology and the open-source platform "Hugging Face" shows that in the past year, the global download share of open-source models developed in China reached 17.1%, surpassing the United States' 15.8% and ranking first globally.
This change is driven by emerging forces represented by China's large models, which have gained greater influence and voice through continuous open-source openness and application advantages in vertical scenarios, competing against Silicon Valley giants like OpenAI and Anthropic who strictly guard their closed-source fortresses and SOTA explanation rights.
International mainstream AI platforms such as Vercel, Kilo Code, Roo Code, Cline, Factory Droid, and Okara, which are closely watched by overseas users, have announced immediate integration with GLM-4.7, reflecting the reality of Chinese large models being favored globally. The U.S. AI inference platform Fireworks called GLM-4.7 "a wonderful holiday gift."
According to reports from foreign media such as the Financial Times of the UK and Wired of the US, the past year was a harvest year for China's open-source large models. At the beginning of the year, DeepSeek R1 established high standards for open-source models with its excellent reasoning capabilities, greatly boosting confidence in the potential of China's domestic open-source technology worldwide. Subsequently, the Tongyi Qianwen series provided developers with rich choices by covering different scales and scenarios. In the middle of the year, GLM-4.5 and Kimi K2 were introduced, achieving breakthroughs in capabilities such as intelligent agents, long context, and complex reasoning through innovative training methods and architectures. At the end of the year, the newly released GLM-4.7 enhanced coding capabilities, long-term task planning, and tool collaboration, immediately ranking first on the "Hugging Face" global trend list.
The influence of China's large models has also accelerated their "going out" efforts, increasingly gaining recognition from the global capital market. In the last month of 2025, Beijing ZhiPu Huazhang Technology and Shanghai XiYu Technology have both passed the Hong Kong Stock Exchange's listing review. Among them, ZhiPu plans to officially list on the main board of the Hong Kong Stock Exchange on January 8, 2026. "This is an important sign of the maturity of China's AI large model industry," said Li Qing, director of the Greater China region of the internationally renowned consulting company Frost & Sullivan. He pointed out that the capital market has begun to give institutional tolerance to AI companies with high R&D investment and core technical barriers but not yet profitable. The Hong Kong stock market has become a key bridge connecting China's AI industry with the global capital market. China's large model industry is moving from early technological exploration into the stage of commercial verification and large-scale implementation, starting a new cycle of global layout and competing with global AI giants on the same stage.
According to Frost & Sullivan's forecast, the market size of China's large models exceeded 20 billion yuan in 2024 and is expected to maintain a compound annual growth rate of more than 40% over the next three years.
Original: toutiao.com/article/7589941831547257354/
Statement: This article represents the views of the author.