Temperature is a setting that changes how much variety the model allows when it picks each next word. Lower means more expected, repeatable wording; higher gives it more room to vary. Turning it up adds no intelligence, and turning it down adds no truthfulness. Most clients never touch it directly. It lives in APIs (developer-level access), custom agents, and advanced settings, and it’s one reason the same prompt won’t give identical output every time.
A script versus a brainstorm is the everyday contrast. Confirming office hours, you want the script. Naming a workshop, you want more range. Same person, different amount of freedom.
Why you care
Match the freedom to the work. Classifying transactions, extracting invoice fields, or following a strict format wants low temperature, the same kind of answer every time; naming ideas and first drafts benefit from more range, and you still review the work. More variation stays prediction rather than real creativity, and a low-temperature answer can still be a hallucination without the right context, so keep human review either way.