A report shows that China’s open source artificial intelligence models account for nearly 30% of the total use of the technology globally, and the use of Chinese prompts ranks second after English.
This year’s surge in global open source large language model (LLM) usage has been driven by systems developed in China, including Alibaba Group Holding’s Qwen series of models, DeepSeek’s V3 and Moonshot AI’s Kimi K2, according to a recent report from third-party AI model aggregator OpenRouter and venture capital firm Andreessen Horowitz. Alibaba owns the South China Morning Post.
Western proprietary models such as OpenAI’s GPT-4o and GPT-5 still dominate, accounting for 70% of the global share.
Do you have questions about the hottest topics and trends from around the world? Get the answers with SCMP Knowledge, a new platform of curated content with explainers, FAQs, analyzes and infographics brought to you by our award-winning team.
According to OpenRouter’s empirical research on 100 trillion tokens, China’s open source LLM global share started from a low base of 1.2% at the end of 2024 and reached nearly 30% within a few months of this year. Tokens are units of data processed by AI models during training and inference, enabling prediction, generation, and inference.
So far this year, with growth accelerating in the second half of 2025, Chinese open source LLMs account for an average of 13% of weekly token trading volume, nearly matching the 13.7% average recorded by AI models in the rest of the world, the report said.
“China has become a major force, not only through domestic consumption but also through the production of globally competitive vehicle models,” the report said.
The report provides new evidence that China has become a rival to the United States in the development of artificial intelligence models, even as Washington restricts Chinese companies’ access to advanced graphics processing units from companies such as Nvidia and Advanced Micro Devices.
Western proprietary large language models, such as OpenAI’s GPT series, continue to lead the global use of artificial intelligence. Photo: Shutterstock alt=Proprietary large Western language models, such as OpenAI’s GPT series, continue to lead the world’s use of artificial intelligence. Photo: Shutterstock>
The rise of open-source AI models in China “reflects not only the quality of competition, but also rapid iterations and intensive release cycles,” the report said.
The report pointed out that Alibaba Cloud Qwen and DeepSeek’s aggressive release schedule allows users to quickly adapt to the increasing development workload. Alibaba Cloud is the artificial intelligence and cloud computing services arm of Hangzhou-based Alibaba.
As China’s open source model gains recognition for improving efficiency and low-cost adoption, Chinese has become the second most used prompt language globally, accounting for nearly 5% of all requests, behind market leader English, the report said.
This is significantly higher than Chinese’s share of the internet, which is around 1.1%, according to various estimates.
The report shows that in terms of global share of LLM tokens, China ranks fourth, behind the United States, Singapore and Germany.
The report states that global demand for open source artificial intelligence models has evolved from a monopoly dominated by DeepSeek in December 2024 to a fragmented landscape with competition between Alibaba Cloud Qwen and Moonshot AI’s Kimi by the end of 2025. The report says LLM use is now widespread, with no single model accounting for more than 25%.
This article originally appeared in the South China Morning Post (SCMP), the most authoritative voice on China and Asia for more than a century. For more SCMP stories, explore the SCMP app or visit SCMP on Facebook and twitter Page. Copyright © 2025 South China Morning Post Publishing Limited. All rights reserved.
Copyright (c) 2025. All rights reserved by South China Morning Post Publishing Limited.