How much are you saving once the more expensive model already has all the context loaded and ready to go?

API calls get more expensive, not less, as you've loaded more context. This is exactly when you want to switch to cheaper models.