Why count them?

Text takes up space in a context window. Tokens are a way to measure how much text a model is handling.

Keep the distinction

Tokens are the units. The context window is the working space. Inference is the model doing the work.