What is temperature for in a language model?
Answer out loud or in your head first, then compare.
Answer
Temperature is one of the parameters that affect how the next tokens are chosen, if the given model exposes it. Simplified, a lower value favours more predictable choices, and a higher one can increase the variety of answers.
It's not a truthfulness slider. A low value won't stop the model making mistakes, and zero doesn't give a general guarantee of an identical result on every run. For data extraction I usually care more about stable behaviour than inventiveness. I still check the result with a schema and quality tests. Not every modern model lets you set this parameter freely.
Why it matters: A common basic question about model APIs.
Possible follow-ups: Why doesn't lower randomness replace checking sources?
How did it go? Be honest — “Not yet” puts the question on your review list.
Further reading
- Claude Platform: Define success criteria and build evaluations(opens in a new tab)Docs; accessed 22.10.2026. Success criteria, representative data and kinds of graders.