Chinese AI models have overtaken US rivals in token consumption since February, according to OpenRouter data, marking a significant shift in the global AI race. DeepSeek and MiniMax are leading this charge, with token usage rising 476% in just one month. This surge reflects both cheaper pricing and more efficient architectures that appeal to cost-conscious developers worldwide.

Chinese companies charge $2 to $3 per million output tokens compared to about $15 for US competitors like Anthropic's Claude Sonnet 4.5 - a near sixfold gap that becomes critical as AI agents consume vastly more tokens than traditional chatbots. Developers like Terry Zhang in Hong Kong now mix Chinese and US models, spending $50 daily on Moonshot's Kimi versus the $900 daily cost of using only Claude. This pricing advantage stems from cheaper energy and efficient 'mixture-of-experts' designs.

The trend extends beyond pricing to revenues, with MiniMax's M2.5 model now ranked among the most-used globally. Alibaba has responded by creating a dedicated Token Hub business unit, signaling its belief that token economics will define the next phase of AI competition. As billions of AI agents emerge as the primary interface between people and digital services, China's cost advantage in token production could reshape global AI adoption patterns.