According to the latest calculations by the third-party aggregation platform OpenRouter: During the statistical week of October 19th to 25th, the total invocation volume of global AI large models reached 188 trillion Tokens, a 14.63% increase over the previous week, and the popularity of commercial implementation is still rising.
Open the plate up to examine it. The models of Chinese manufacturers used a total of 93.54 trillion Tokens last week, a 18.14% increase over the previous week; the American models used 22.47 trillion Tokens, with a 29.44% increase over the previous week, but their scale is completely different from that of China’s models. This marks the twenty-sixth consecutive week where the Chinese large-scale models have exceeded those of the United States in weekly usage, taking the top position globally. The widespread application of AI in both B and C scenarios in China has given it a clear scale advantage in overseas markets.
More iconic are the positions. Last week, all five spots in the global call volume ranking were occupied by Chinese AI large models, with no overseas leading product making it into the top five. This is rare in the history of the rankings. Behind this gap in scale lies the activity levels on the application side: domestic models are widely used in high-frequency scenarios such as office work, customer service, and creative tasks, which has boosted call volumes; while overseas products’ growth comes more from higher-priced enterprise deployments, with each transaction being valuable but a total amount that remains difficult to match.
The model that reached the top is V4.1-Flash-1017, with a single-week invocation volume of 33.6 trillion Tokens. The value increased by 801% compared to the previous week—making it the most rapidly growing popular model globally recently. A single product nearly consumes nearly one-fifth of the total global invocations, indicating the degree of concentration among leading models.