The VS code open AI chat extension sends [the file that is currently open in the editor] along with the prompt.

In the chat prompt input box where it says, "Ask Copilot". There is a button that says "Add Context" and, next to that, the currently-opened file. Either click it to disable the current file context, reduce the file size in the editor, or choose a less-expensive model to ask in the chat box.

The crux of it is, if you're editing a large file, it will consume a lot of tokens.