I am utilizing the identical system prompt and user data for GPT to produce a descriptive analysis. However, upon making subsequent calls to the OpenAI API, I observe variations in the results. For instance, in the initial request, the analysis indicates that 40% of people favor the product, but in the subsequent run, this figure shifts to 20%. Additionally, I’ve observed that making frequent API calls yields little consistent results, whereas hitting the same API after a few days produces different outcomes.