Temperature explained — A setting that affects sampling randomness
For models that support it, temperature changes how the next output is sampled. Lower settings generally concentrate choices; higher settings allow more variation.
An everyday example
You compare varied slogan ideas, while a fixed-format extraction task prioritizes consistency and validation.
When you’ll encounter it
Adjust it only when the tool exposes the setting and the model supports it, then compare results.
What to watch for
It is not an accuracy switch. Zero does not guarantee identical or correct answers. Supported ranges, defaults and configurability vary.
Related terms & sources
Prepared from the public sources below. Examples illustrate the concept. Reviewed on 2026-10-15.