When I ask Grok the same question in Auto and Think modes, I sometimes get noticeably different answers—not just a longer explanation. Is Think using a different reasoning process, or is Auto choosing a different model or amount of effort based on the prompt?

I’m curious how other people decide which mode to use, especially for questions where accuracy matters more than response speed.