I said something similar here.
The above workflow took, by memory, about 4 tokens using Pro+.
Copilot Chat uses one premium request per user prompt, multiplied by the model's rate. This includes ask, edit, agent, and plan modes in Copilot Chat in an IDE.