Clarification on token pricing for multiple completions (n>1) in a single API call"

API
1

I’m working on a project that requires multiple completions for a single prompt using the OpenAI API. I’m trying to understand how the token pricing works when requesting multiple completions (n>1) in a single API call. Specifically:

Are input (prompt) tokens charged once or multiple times based on the number of completions requested?

I’ve looked through the documentation but couldn’t find clear information on this specific topic. The documentation does say, “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.”

1 Like
2

This is a relatively easy question to answer.

GPT-4-turbo-2024-04-09:

“usage”: {
“prompt_tokens”: 211,
“completion_tokens”: 100,
“total_tokens”: 311
},

GPT-4-turbo-2024-04-09 @ n=10:

“usage”: {
“prompt_tokens”: 211,
“completion_tokens”: 1000,
“total_tokens”: 1211
},

In this case, sending with top_p:0.001, n>1 is of little value, because it provides almost identical answers…

1 Like