OpenAI can respond with multiple responses after sending one prompt. Example would be - I have a large large dialog, lots of tokens in use, and I am looking for a specific code solutions. But if gpt gives me poor code - I need to rerun it again and spend all that money on the prompt again. Or I can use something like n=3 and have 3 responses to try for the price of one prompt. Alternative use - if I need 10 ads written, Id rather use n=10 instead of asking for 10 ads in one response, because every next one is influenced by the previous one.