Why do you charge an $0.25 per million tokens regardless of whether they are write read or cached?
This fee creates extreme markups on usage. Example:
GPT5.1 Codex Mini:
37.8M tokens (36.2M cache read, 1.5M input, 0.06 M output)
Cost: $10.87 on teams plan ($9.45 cursor token fee)
Cost: $1.45 on any other plan.
Please explain the logic here.