SmartBrain AI's overseas platform has been exposed to abnormal caching billing, with the Token deduction amount enlarged by nearly 19 times

According to DynaInsight Beating monitoring, the Z.ai overseas developer platform of Intelliburst AI has been exposed to a serious cache billing system anomaly. Developer Da7em reported to the official team that when using the GLM model on the platform, the repeated context seemed unable to trigger the cache deduction correctly, instead being treated as a completely new input for duplicate billing.

The billing screenshot shared by Da7em indicates that the actual Token usage recorded by the Zcode development tool was about 270,000, but the platform's final billing statement showed a consumption of nearly 5 million Tokens, magnifying the bill by almost 19 times. To investigate any client-side interference, Da7em compared and tested multiple intelligent agent development frameworks such as Claude Code, Hermes Agent, Zcode, and OpenCode, confirming consistent anomaly behavior across different tools, thus identifying the root cause of the issue in the server-side cache and accounting system.