Building a solution based on GPT3 API and would like to make cost prediction.
However, the following point is not clear from the documentation and different API vendors approach this differently, so I want to be sure to plan soland searching the forum didn’t help either.
For example, 1000 tokens for GPT-3 davinci-003 model costs $0.02.
So if my API request (prompt request + model response) contains, say, 100 tokens, I will be charged $0.002 (for 100 tokens only), or it will be charged $0.02 as for 1000 tokens batch (which I did not use however)?
I’m not an authority but I have interpreted it as 100 tokens will cost 10% of the cost of 1000. And while I have not done the maths, looking at my billing history suggests that’s what’s happening.