Today, the Chinese Government Website forwarded an article from the People's Daily, quoting the National Data Administration, officially naming the fundamental unit of large models "Token" as "Word Element" in the official context. The report revealed that the daily usage of Word Elements in China's artificial intelligence industry has exceeded 140 trillion times. This authoritative naming not only ends the long-standing debate in academia and industry on the Chinese translation of Token, but also marks further standardization of the technical terminology for large model technology at the national level.
Previously, there were various competing proposals for the name of Token within the industry. Yang Bin, a scholar from Tencent Research Institute, proposed "Model Element," which sparked widespread discussion. Later, the "Intelligence Element" proposal recommended by Wang Xiaochuan, founder of Baichuan Intelligence, You Yang, founder of Lunchen Technology, and Hu Yilin, former associate professor at Tsinghua University, gained high visibility on social platforms due to its emphasis on the characteristics of Token as a "general intelligent" unit rather than just a "general computing" unit.
However, the official term "Word Element" chosen by the government focuses more on the technical underlying logic. According to the official explanation, a Word Element is the smallest unit for AI to understand human language, with a segmentation granularity between characters and words, allowing more accurate coverage of semantic slices during text processing by large models. As the usage volume approaches the historic milestone of 140 trillion, a unified terminology standard will help reduce industry communication costs and lay a solid foundation for the statistical quantification and policy-making of subsequent AI infrastructure.
