Chinese AI models have staged a silent revolution and now account for nearly 30% of global open-source AI usage. It is a dramatic rise that reflects a major shift in the technology landscape, as Chinese-developed systems challenge the Western dominance despite ongoing restrictions on advanced chip access.
A study estimated to have analyzed 100 trillion tokens puts into great detail how quickly Chinese AI models gained ground. Chinese open-source large language models, starting from just 1.2% of global usage in late 2024, skyrocketed to nearly 30% market share in a matter of months, according to research from OpenRouter, a third-party AI model aggregator, and venture capital firm Andreessen Horowitz.
Among them are models developed by major Chinese tech companies: Alibaba Group’s Qwen family of models, V3 of DeepSeek, and Moonshot AI’s Kimi K2, among others, have come out as globally competitive alternatives. These systems have attracted users worldwide through a combination of efficiency, cost-effectiveness, and rapid innovation.
That growth has been especially explosive in the second half of 2025: Chinese open-source models now average 13% of weekly token volume, nearly matching the 13.7% average recorded by AI models from the rest of the world combined. Tokens are fundamental units of data that AI models process during training and when generating responses.
With systems such as OpenAI’s GPT-4 and GPT-5, Western proprietary models still have overall dominance, holding roughly 70% of the global market. Nevertheless, the velocity at which China is advancing has taken many observers by surprise.
What makes China’s progress particularly noteworthy is the context: Washington has imposed significant restrictions preventing Chinese firms from accessing advanced graphics processing units from manufacturers like Nvidia and Advanced Micro Devices. These chips are considered vital for training the most advanced AI models. Yet Chinese developers have found ways to compete anyway.
DeepSeek and the Competitive Ascent of China’s AI Open-Source Models
According to the report, a number of factors are helping China succeed: Chinese companies have been able to adopt aggressive release cycles, churning out new model versions and updates much more rapidly than rivals. The “dense release cycle” allows users to adapt rapidly to improvements and keeps the Chinese models competitive with Western alternatives.
In this regard, Alibaba Cloud’s Qwen and DeepSeek have been very active, releasing enhanced versions from time to time to address user feedback and expand capabilities. This strategy has helped build loyal user bases both inside and outside China.
Cost efficiency has also played an important role. Chinese open-source models often provide similar performance compared to Western alternatives at significantly lower operational costs. This value proposition has been compelling for businesses and developers working on shoestring budgets.
One interesting linguistic effect of the global adoption of Chinese AI models is that Chinese has become the second most-used global prompt language for interacting with AIs, with almost 5% of all requests. Only English ranks higher. This percentage notably outpaces the general internet presence of the Chinese language, at an estimated 1.1%, according to various estimates.
The competitive landscape has been evolving rapidly. Where DeepSeek held a monopoly position, of sorts, among open-source models in December 2024, by late 2025, the market is fragmented. Competition from Qwen by Alibaba Cloud and Kimi from Moonshot AI has distributed usage more widely, with no single model now exceeding 25% market share among open-source options.
Emerging as a Global AI Peer to the US Despite Tech Restrictions
Geographically, China ranks fourth globally in terms of large language model token volume, while the United States ranks first, followed by Singapore and Germany. However, the gap seems to be getting smaller as Chinese domestic usage grows and international adoption increases.
The findings offer new evidence that China has emerged as a true peer to the United States in the development of AI, despite technology restrictions. Chinese companies have proven they can create globally competitive models that users actively opt for over alternatives.
Industry observers say this represents not only technological achievement but reflects the strategic priority placed by China on the development of AI. The government has emphasized that artificial intelligence is a key development area, with heavy investment in research and development.
Given the continually moving landscape of AI, Chinese and Western models will also likely continue to compete with one another. Chinese developers have proven themselves capable of innovating quickly and delivering compelling products. Whether they will be able to sustain this in a geopolitically complex environment is yet to be seen, but their presence in the global AI ecosystem can no longer be ignored.
