Chinese AI models gain global users

China is stepping up its push for a more open, inclusive and collaborative approach toward artificial intelligence, as the country's open-source AI ecosystem gains growing recognition among overseas users with its wider and fairer access to high-performance AI technologies at lower costs, said senior officials and industry experts.

"China is ready to work with all parties to establish a globally interoperable AI governance system that accommodates the interests of all parties and promotes beneficial, safe and fair AI development," said Li Lecheng, minister of industry and information technology, when attending the first session of the United Nations Global Dialogue on AI Governance last week in Geneva, Switzerland.

By placing great importance on both AI development and governance, the nation has supported the world's first resolution on AI standardization and has worked with international partners on more than 120 AI standards, contributing practical experience to the global intelligent transformation, Li said.

He added that China will continue to help bridge the global AI divide by supporting fair and inclusive development and by assisting countries, particularly those in the Global South, in building AI technologies and services.

Mathias Gabrysch, senior director of ecosystem development at global communications alliance GTI, noted during the conference that China has provided "valuable practical experience for the world" by making extensive AI applications in agriculture, healthcare and public services.

"More important, China's open-source models — such as DeepSeek — enable researchers, startups and public institutions around the world to access high-performance AI at low cost. That is a genuine contribution to the global public good," he said.

Speaking to China Daily more recently, Pan Helin, a member of the Ministry of Industry and Information Technology's Expert Committee for Information and Communication Economy, said the growing international appeal of Chinese AI models is being driven by their improving capabilities, competitive pricing and open-source nature.

"The gap between AI models made by China and the United States is narrowing rapidly. China's significant cost advantage is increasingly offsetting the remaining difference in performance. In some areas, such as video-generation models, Chinese developers are already ahead of their overseas competitors," Pan added.

According to OpenRouter, a platform widely used by developers to access AI models, weekly application programming interface calls for Chinese large language models have surpassed those for US-developed LLMs for 11 consecutive weeks as of Sunday, placing China at the top position globally.

Since Feb 8, more than 30 percent of tokens used by US companies on the platform each week were related to Chinese AI models, with the share peaking at 46 percent. Tokens are small units of data processed by AI models during training and inference. In contrast, the share was only 4.5 percent in the first half of 2025.

"Whether individuals or businesses, users worldwide are validating China's AI development path through real-world adoption. That, in turn, will further strengthen the global application ecosystem for Chinese AI models," Pan said.

A growing number of overseas companies are already turning to Chinese LLMs as a more cost-effective option for deploying advanced AI.

In late June, Brian Armstrong, CEO of cryptocurrency platform Coinbase, shared on social media that the company had added two Chinese LLMs as default models for its engineering team, cutting AI spending to nearly half its peak level.

The same month, AI startup Lindy announced that it had moved all its traffic from Anthropic's Claude to DeepSeek-V4.

"(This) saves us millions of dollars and we are actually seeing an increase in performance on many core use cases. (It is) transformative for the business," Lindy CEO Flo Crivello wrote on social media.