Record the Model Alongside the Prompt

A saved prompt is not a complete experiment record. The model, relevant settings and test inputs can all affect the result.

When comparing an OpenAI workflow, record the exact model identifier used, the prompt version and the evaluation set. Keep the response and acceptance result for each case. If you choose a dated model snapshot where supported, still plan how you will evaluate a later migration.

Avoid changing several settings at once and then attributing the improvement to one of them. Start with a clear baseline and inspect the differences. This record is useful when an answer changes unexpectedly: you can investigate which input or configuration moved instead of relying on a screenshot of an earlier successful response.

OpenAI documentation.