[Zhipu AI Overseas Platform Exposed for Billing Anomaly, Token Charges Amplified Nearly 19 Times] According to monitoring by Beating, the Zhipu AI overseas developer platform Z.ai has been exposed for a severe anomaly in its caching billing system. Developer Da7em reported the issue to the official team on X, stating that when using the platform's GLM model, repeated context inputs seemingly failed to trigger proper cache deductions and were instead treated as entirely new inputs, resulting in duplicate charges. A billing screenshot disclosed by Da7em shows that the actual Token usage recorded by the development tool Zcode was approximately 270,000, but the platform's final billing statement indicated a consumption of nearly 5 million Tokens, amplifying the billing amount by nearly 19 times. To rule out client-side interference, Da7em conducted comparative tests using multiple intelligent development frameworks, including Claude Code, Hermes Agent, Zcode, and OpenCode, confirming that the anomaly was consistent across different tools. Therefore, the root cause was determined to lie in the server-side caching and accounting system. [Original Link]