On November 2, according to the latest data estimates released by OpenRouter, an open-source large-model aggregation platform, in the past week (October 26 to November 1), the total invocation volume of global AI large models reached 194 trillion tokens, a steady week-over-week increase of 3.19%. In this global competition across the AI computing-power and model-application ecosystem, Chinese large models showed extremely strong growth momentum.
Specifically, the total weekly invocation volume of the listed Chinese AI large models soared to 99.55 trillion tokens, up 6.43% week over week. By contrast, the weekly invocation volume of U.S. AI large models during the same period was 20.68 trillion tokens, down 7.97% week over week. This also marked the 27th consecutive week that the weekly invocation volume of Chinese large models surpassed that of the U.S., keeping China firmly in first place globally.
In the breakdown rankings, among the top five global AI large models by total invocation volume, China's homegrown large models strongly took four spots. Among them, DeepSeek V4.1Flash1017 remained in first place, demonstrating extremely high market recognition and adoption. Meanwhile, MiniMax M3.1Flash also saw explosive growth this week, rising to eighth place on the list, with its weekly invocation volume surging 690% week over week, making it another standout rising star in the large-model application ecosystem.
