Responses Affected by Previous Prompts?

Hi, is it possible that in one session in case of tens or hundreds of repetitions of the same task, each time with different data, 4o-mini responses start getting less accurate? I am using it to rank something and the ranking seems to get worse after some iterations, so I am investigating the cause.

Currently the API is stateless. Therefore, technically there can’t be a link between two API calls unless you are including previous responses.