#Why are the results I get from the chatgpt api noticeably worse than when I use the webapp?
5 messages · Page 1 of 1 (latest)
Temperature would make it provide a different(potentially worse) reply. I don't know enough about your document to judge the 2 examples that you provided. For me, both of them don't make sense(likely due to not knowing the context).
Temperature absolutely relates to output quality.
Hyperparameters.
hi, gpt-3.5 model is ideal for like mathmatic staff, to get detailed information from your question, I suggest to try using text-davinci-003 or gpt-4