How Token Economics Aggregates Compute, Models, and Corpora: Measuring Intelligence Through a Single Robotic Grasp
AI summary
As of June 2026, China's average daily token call volume has exceeded 500 trillion, and Token is becoming the metering interface connecting model services, inference compute, and data supply. DeepSeek has explored peak-valley time-of-use pricing since August and adjusted prices for its Flash series models on September 10, with the off-peak cached-hit input price at 0.02 yuan per million tokens. Infinigence AI disclosed that its cross-cluster heterogeneous inference architecture, under specific test conditions, reduced time-to-first-token by 51.5% and cost per token by 37.5%.
Body
The English body is being prepared. You can open the original source.