I’m using GPT4 to translate lines of text. If I have a long document, is it cheaper to send the whole thing and just wait for GPT to finish translating it or to send a bunch of requests translating it a couple of sentences at a time?
Does it matter how many requests you send OpenAI pricing wise or is payment just based on amount of text you send, not how many requests you send?
Assuming use of the API, you pay for tokens used, prompt tokens are half the price of completion tokens, you can think of that as input is half the price of output. So it would make no difference if you were to do a paragraph at a time or a page at a time. You may gain some additional translation context from a page at a time as there is more text to use as reference.
You may also gain a better quality translation by using two passes, one to translate and one to tell the AI that the text is translated and to reconstruct the text as a native speaker of the language would, clearly, at double the token cost.