What does one AI message actually cost?
One message costs the input tokens times the model's input price, plus the output tokens times the output price. A short reply on a small open model is often a fraction of a cent. A long reply on a large model can be several cents.
Meterwise estimates the worst case before the call, using your output limit, then replaces that hold with the token counts the provider returns. The price table is dated and has to be checked by the operator.