Artificial intelligence Photo: VCG

The US-based OpenRouter, a leading global AI platform, released its latest weekly ranking of AI model usage, with all of the top five models developed by Chinese companies, a trend that a Chinese industry analyst said underscored the rapid rise of Chinese tech firms in the emerging AI sector despite the US' attempts to contain their development.

According to OpenRouter's latest weekly AI model usage rankings, the top five models were all developed by Chinese companies. DeepSeek's V4 Flash topped the leaderboard with 7.22 trillion tokens of usage, followed by Xiaomi's MiMo V2.5, Tencent's Hy3, DeepSeek's V4 Pro and Z.ai's GLM 5.2.

The rankings measure total token usage across platform, with token counts reflecting both prompt and completion tokens processed by each model.

Beyond their cost-effectiveness, the top five Chinese models on the ranking each demonstrate distinct strengths. The second- and fifth-ranked models were both developed by DeepSeek, with the two offering complementary capabilities to meet different development needs. Its flagship Pro version delivers performance comparable to leading overseas closed-source models in areas such as coding and complex AI agent tasks, China Newsweek reported. 

A common feature shared by these models is that they quickly climbed the rankings shortly after their release. DeepSeek's V4 Flash officially opened API public testing on Friday, while Tencent's Hy3, which ranked third, was only officially open-sourced on July 6.

Also, on Hugging Face, the world's leading open-source AI community, open-source models developed by Chinese companies have recorded the highest cumulative download volume globally. Kimi-K3 ranked first with 968,000 downloads and 9,770 likes, followed closely by DeepSeek V4 Flash.

The latest rankings show that Chinese AI large language models are rapidly expanding into global markets and gaining broader recognition among developers and users, Xiang Ligang, director-general of the Zhongguancun Modern Information Consumer Application Industry Technology Alliance, told the Global Times on Monday.

Despite the US' ramp-up attempts to curb the development of China's AI industry through technological restrictions, export controls and other measures, they have failed to slow China's pace of AI innovation, the expert said.

AI competition is not merely a contest of individual technologies or chip capabilities, but a comprehensive competition involving algorithms, data, computing power, application ecosystems and talent pipelines. Xiang said that China has vast data resources, a complete manufacturing ecosystem and abundant application scenarios, enabling large models to be rapidly deployed across sectors including finance, manufacturing, transportation and healthcare. 

By the end of 2025, China had more than 6,000 AI companies, with the scale of its core AI industry expected to surpass 1.2 trillion yuan, up nearly 30 percent year-on-year, the Xinhua News Agency reported.

The cumulative global downloads of China-developed open-source AI models exceeded 10 billion, while China became the world's largest holder of AI patents, accounting for 60 percent of the global total, according to Xinhua.

Meanwhile, Chinese companies have actively pursued open-source approaches, lowering the barriers to technology adoption, attracting global developers to participate and further enhancing the international influence of Chinese AI models, Xiang said.

Certain external restrictions may create short-term hurdles, but they have also accelerated Chinese companies' pursuit of technological self-reliance and fostered a more complete AI industrial chain, with an open innovation ecosystem, Xiang said.