On October 19, data sourced from OpenRouter revealed that the total global invocation volume of AI large models for the week of October 12 to 18 reached 164 trillion Tokens, marking a 4.65% decrease from the previous week. Specifically, China's weekly invocation volume for AI large models stood at 79.18 trillion Tokens, experiencing an 8.72% decline week-over-week. In contrast, the U.S. recorded a weekly invocation volume of 17.36 trillion Tokens, down by 8.49% compared to the preceding week. Notably, China has maintained its lead over the U.S. in weekly AI large model invocations for an impressive 25 consecutive weeks, securing its position as the global frontrunner.

A striking observation from the data is that the top five AI large models with the highest invocation volumes worldwide last week were all developed in China. Among them, DeepSeekV4.1Flash emerged as the leader for the second consecutive week, with a weekly invocation volume of 32.8 trillion Tokens, despite a 20% week-over-week decrease. Zhipu GLM5.3Flash claimed the second spot, registering a weekly invocation volume of 12.2 trillion Tokens. Meanwhile, Step5Preview from StepFun ascended to third place, boasting a weekly invocation volume of 11.5 trillion Tokens, which represented a significant 30% increase from the previous week.