China's daily average AI token consumption reached approximately 500 trillion in June 2026, nearly four times the level recorded in December 2025, according to a joint report by the China Academy of Information and Communications Technology (CAICT) and other institutions. The "Large Model Token Service Performance Monitoring Report (2026)" also revealed that domestic models achieved an average throughput of 78.23 tokens per second in August 2026, doubling the 2025 average of 38.25 tokens per second but still trailing typical foreign models at 112.82 tokens per second. The report compared performance across DeepSeek, Qwen, GLM, and Kimi. On pricing, the normalized average cost of domestic model services rose to 13.46 yuan since 2026, nearly doubling year-over-year, yet remaining far below the foreign average of 117.8 yuan.