China's AI Model API Call Volume Surpasses US for Six Straight Weeks — MiniMax M3 Breaks Into Top Three Globally

For six consecutive weeks, Chinese AI models have outpaced their American counterparts in total API call volume on OpenRouter, one of the largest AI model routing platforms, signaling a decisive shift in global developer adoption patterns. Data from OpenRouter reveals that in the most recent tracking period, total weekly global API calls reached 36.1 trillion tokens. Of that, Chinese models accounted for 14.19 trillion tokens, compared to just 3.2 trillion tokens for US models — a commanding lead that has held steady for over a month. Leading the charge is DeepSeek-V4-Flash, which maintained the No. 1 spot on OpenRouter's rankings by API call volume. The model, known for its strong reasoning capabilities and dramatically lower pricing compared to frontier US models, has become a favorite among developers building agentic workflows and high-volume production applications. In second place sits Tencent Hy3 preview, Tencent's Mixture-of-Experts architecture model with 295 billion parameters and a 256K context window. Its open-source release and competitive pricing have driven rapid adoption since launch. The most striking development this week is MiniMax M3 breaking into the top three globally for the first time. The model, developed by the Shanghai-based AI startup, climbed rapidly through the ranks as developers discovered its strong performance-to-cost ratio. Xiaomi also entered the rankings with MiMo-V2.5 securing the No. 4 position, marking the smartphone giant's growing ambitions in foundational AI. Meanwhile, Anthropic's Claude models have suffered a notable decline. Both Claude Sonnet 4.6 and Claude Opus 4.7 dropped out of the top five entirely — a significant fall from grace for models once considered among the most capable on the market. Industry observers attribute the decline to their relatively higher per-token pricing and the increasing availability of Chinese alternatives offering comparable performance at a fraction of the cost. This shakeup reflects a broader trend: developers are increasingly prioritizing cost efficiency over raw benchmark performance. As AI usage scales from experiments to production deployments, the total cost of API calls has become a decisive factor in model selection. Chinese AI companies, backed by aggressive pricing strategies and rapidly improving model quality, are well-positioned to capture this demand. With no signs of the momentum slowing, Chinese AI models appear poised to extend their dominance on OpenRouter in the weeks ahead.