China’s AI Models Process 98 Trillion Tokens, 85% Above US

China now accounts for roughly 40% of the 50 most widely used artificial intelligence (AI) models worldwide. Its models have also overtaken US rivals in monthly token usage.

The numbers point to developers increasingly reaching for Chinese systems, even as US usage keeps rising.

China Closes the Gap in Global AI Usage

Just 5 of the top 50 models were Chinese at the start of 2025. That count reached around 20 by May 2026, a fourfold rise that lifted its share to 40%.

According to Apollo Global Management, US models moved in the opposite direction. Their count in the top 50 fell to around 28 from 33 over the same stretch.

Token usage tells a sharper story. Among the top 20 most used models, Chinese AI models processed 98 trillion tokens in June, against 53 trillion for US models.

Chinese token usage climbed 113% from May to June. US usage grew 43% over the same month. That widened China’s lead to 85%, well above the 24% margin recorded in May, the Kobeissi Letter noted.

Rising Usage Meets Deepening US-China Friction

The usage boom arrives as ties between US and Chinese AI firms fracture. Alibaba banned its staff from Anthropic’s Claude Code from July 10, citing back-door risks, and told them to switch to its own Qoder tool.

Anthropic had earlier accused Alibaba of the largest known distillation attack against it. The firm has also pressed Washington for tighter chip export controls, arguing the US can still secure a decisive edge.

“But if the US and its allies act now to address both issues, it may be possible to lock in a 12-24 month lead in frontier capabilities. A lead that large by 2028 would be enormously advantageous,” it said.

Beijing is tightening its own grip. A regulatory purge removed more than 14,000 non-compliant AI products from Chinese networks this month.

Subscribe to our YouTube channel to watch leaders and journalists provide expert insights