New: Cookbooks and AI ExplanationsStep-by-Step recipes to solve problems connected to Roadmaps and Cheat Sheets. Need more details? Use AI buttons for structured and simple explanations with concrete examples throughout the whole platform.Take a look
Temperature zero, same prompt, two different answers.
What you'll have at the end
A logged set of five repeated runs at temperature zero, each marked identical or different
You need
A prompt you can send to a model repeatedly with temperature fixed at zero and no memory of earlier turns between sends, through a raw API call, a developer console, or a chat session you restart each time.
Not covered
Building a fully reproducible local inference setup of your own is a separate, much bigger project; this only measures and narrows the drift in whatever hosted endpoint you already call.
Turn the temperature down so the same prompt gives a repeatable answer
start there first if you have not set temperature to zero at all yet; this recipe picks up once you already have and still see drift
Match temperature and top_p to the job instead of the defaults
go there once you need a non-zero temperature on purpose for a creative task, instead of leaving the default in place
Checked 15 Aug 2026
Part of the Large Language Models (LLMs) cookbook