0
I’m testing the same prompt several times through the OpenAI API, but the wording and sometimes the details change between responses. The requests appear identical on my end.
How much of this variation comes from sampling settings such as temperature, and how much can come from model updates or differences in the surrounding request? If I need more consistent outputs, which parts of the request should I check first?
