In plain language
Low temperature is for predictable tasks such as extraction; higher temperature is for brainstorming or creative writing.
Example
A data-extraction step runs at temperature 0 so the same email produces the same fields each time.
Related terms
Large Language Model (LLM) · Token (AI) · Structured Output · Prompt
Common misconception
Temperature 0 does not guarantee identical output every time, and it does not make answers accurate.
Implementation guidance
Use low temperature for extraction and classification, higher for drafting ideas, and test consistency on repeated runs.