Last week, the total number of inference calls for AI large-scale models worldwide reached 188 trillion tokens, marking a week-on-week increase of 14.63%. Among these, Chinese AI large-scale models accounted for 93.54 trillion tokens in weekly inference calls, reflecting an 18.14% week-on-week surge. This marks the 26th consecutive week that China has outperformed the United States, securing the top spot globally. During the same period, AI large-scale models in the United States recorded 22.47 trillion tokens in weekly inference calls, up by 29.44% from the previous week.
Last week, the top five AI large-scale models in terms of global inference calls all hailed from China. The newly launched DeepSeek V4.1-Flash-1017 led the pack, with 33.6 trillion tokens in weekly inference calls, representing a staggering 801% week-on-week increase. The other four models were DeepSeek V4.1 Flash, the newly listed MiniMax M3.1 Flash (free), Zhipu GLM 5.3 Flash, and Tencent Hy4 preview. Additionally, Xiaomi's MiMo-V2.6-Flash slipped to eighth place, while Zhipu GLM 5.3, previously ranked eighth, dropped off the list. These newly introduced models have demonstrated varying degrees of improvement in both performance and cost-effectiveness.