China launches token service climbing plan with 10 tech firms

China's Academy of Information and Communications Technology has teamed up with 10 companies, including Huawei Cloud and Ant Digital, to launch a "Token Service Climbing Plan." The first batch of participants has met the baseline for enterprise-level token service performance: at least 55 tokens per second output, first token delay under 0.9 seconds, and 99.9% call success rate.

Token is the smallest unit for processing text, code, and images in large models, and is becoming the billing unit for AI services. By March 2026, China's daily token calls had exceeded 140 trillion, a thousand-fold increase from 100 billion in early 2024.