With the explosive growth of artificial intelligence technology, Token, as the core unit for measuring and settling large models, is also experiencing an explosive growth. Latest official statistics show that by March 2026, the daily average token call volume in China has surged to over 140 trillion. This astonishing figure not only represents a thousandfold increase from the beginning of 2024, but also saw an increase of more than 40% compared to the end of 2025, highlighting that the application of large models in China is currently in a period of rapid expansion.
First Platform Monitors Large Model Throughput and Latency
To provide objective reference for the growing industrial demand, the Institute of Artificial Intelligence at the China Academy of Information and Communications Technology and other institutions announced that they will hold the "High-quality Token Service Symposium" in Beijing on June 16th. At this industry event, the official will launch the new version of the "Public Cloud Large Model Token Service Performance Monitoring Platform" and release authoritative monitoring reports for the first time. The platform will conduct an objective quantitative evaluation of key performance metrics such as token throughput and latency of current mainstream large model service platforms.
The upcoming "Token Service" series standards will set clear technical benefit boundaries for the underlying computing power and operation capacity services of artificial intelligence in China for the first time. During the forum, invited representatives from domestic top research institutions, major large model manufacturers, operators, and application parties will jointly discuss how to build a more efficient and cost-effective token service ecosystem through in-depth theme sharing.
Launch Special Plan to Empower Trustworthy AI
In addition to releasing performance ranking lists, the conference will also establish a specialized "High-quality Token Service Special Research Group" and launch the "High-quality Token Service Capability Climbing Plan" simultaneously. These measures aim to gather core industry forces and accelerate the development of public cloud large model services in China toward higher quality and greater stability.
Notably, during the conference, an authoritative certification ceremony will be held to issue official certificates to outstanding units that have passed the "Trusted AI - High-quality Token Service Evaluation". Through standard interpretation and the demonstration effect of leading enterprises, the CAICT hopes to guide the entire large model industry to overcome performance bottlenecks and provide a stronger intelligent infrastructure support for the digital transformation of various industries.
