While both Question Answering and CoPilot provides responses to user questions, there are major differences under the hood.

For Question Answering endpoint, the underlying language model is trained to provide the most factual answer based on the data available.

If accuracy is a requirement, Question Answering is recommend.

For CoPilot endpoint, the underlying model assumes that a human is actively checking the result, and thus is will return as much information as possible that is relevant for the response.

If accuracy is not a requirement, CoPilot is generally recommend.