Chinese Language Models Gain Ground in Global Markets

Chinese large language models, abbreviated as LLMs, are gaining significant traction in the global market, boasting 57% of global token usage. This impressive achievement underscores the robust capabilities of China's software and technological sectors, effectively reshaping the landscape of artificial intelligence and international business.

According to a report by the Xinhua News Agency, three Chinese language models successfully maintained their positions within the top five on the OpenRouter leaderboard in terms of global token consumption. OpenRouter serves as a prominent platform for model aggregation and routing, providing essential insights into global trends related to large model utilization.

In AI terminology, a token is the smallest unit of text processed and interpreted by a language model, which can include entire words, partial words, punctuation marks, or even individual characters. The recent data indicates that Chinese AI models have outperformed their US counterparts in weekly token usage for three consecutive weeks.

From March 16 to March 22, Chinese AI models collectively accounted for a staggering 7.359 trillion tokens used, marking a 56.9% increase from the week prior. This surge illustrates not only the popularity of these models but also their expanding influence in worldwide applications.

The global AI community took note of a pivotal moment when Elon Musk commended Kimi, a product of Moonshot AI, for its recent advancement in enhancing Transformer architecture. Such breakthroughs reflect the rapid evolution and competitive edge of Chinese LLMs in the global AI arena.

Market analysts have reported that the overseas revenue generated by Chinese AI companies has now eclipsed their domestic earnings, particularly following the launch of K2.5 AI LLMs in January, showcasing their growing capability to compete on the international stage.

The expansion of Chinese LLMs such as MiniMax, DeepSeek, and Zhipu, backed by major tech firms like ByteDance, Tencent, and Alibaba, signals a strategic leap towards global market integration. Observers note a transformative shift in the utilization of Chinese models abroad, with programming and technical tasks now comprising 39% of their overall usage.

Yan Yijun, vice president of MiniMax, highlighted that a diverse range of users, not just programmers, are now adopting these AI models in their daily routines and workflows. The consistent top ranking of the MiniMax M2.5 model in OpenRouter usage is a testament to this trend.

Insights from MiniMax's 2025 performance report revealed that over 70% of the company's revenue will stem from international markets, reflecting a burgeoning global user base and signaling a shift in the landscape of AI commercialization.

Industry experts, like Tian Feng, underscore that the competitive nature of the Chinese AI ecosystem drives companies to seek opportunities abroad to sustain profitability amid escalating domestic costs. The globalization of these models is predicted to foster explosive growth and redress the balance of technological leadership on the world stage.

As the gap narrows between Chinese LLMs and their international counterparts in terms of programming, multimodality, and reasoning capabilities, the increasing affordability of these models enhances their appeal to foreign users, further cementing their role in reshaping the global AI market.