Usage limits are/were higher in Copilot. They also charge per prompt, not per token.
It is basically a token based pricing, but you get alos a limitation of prompts (you can't just randomly ask questions to models, you have to optimize to make them do the most work for e.g. hour(s) without you replying - or ask them to use the question tool).