When can I expect multiple "choices" in an OpenAI response?

Question

For a simple OpenAI api call in python

response = openai.ChatCompletion.create(
    model="gpt-4",
    messages=[
        {"role": "system", "content": instructions},
        {"role": "user", "content": my_input}
    ])

I get meaningful data in response['choices'][0]['message']['content']

Under what circumstances will there be multiple entries in response['choices']? How should I interpret this?

Answer 1

From what I see, the documentation says that if you set the parameter n to more than 1 you will get more than one "choice".

https://platform.openai.com/docs/api-reference/chat/create

n integer or null Optional Defaults to 1

How many chat completion choices to generate for each input message. Note that you will be charged based on the number of generated tokens across all of the choices. Keep n as 1 to minimize costs.

I don't know the concept behind naming it choices, but I can tell you that the default is 1, so you always will have to access index 0 from choices unless you change the n parameter.