I’m trying to make a small evaluation set where I can compare prompt changes without unrelated output variation getting in the way. If I send the same request to the same OpenAI model with identical settings and a fixed seed, should I expect the response to match exactly?

I’m curious how much repeatability people actually see across repeated calls, and whether model updates or other API-side changes can still affect the result.