TL;DR Do ChatGPT questions asked at the OpenAI playground cost the same as questions sent by my web app to the Open AI web API?
Hi all. Yesterday I used the OpenAI playground to submit ask a particular question to ChatGPT (gpt-3.5-turbo) and obtain a response. The costs as reported by the usage page were 178 total tokens, and $0.000271.
Let’s say I ask the exact same question to ChatGPT (using the same model), and it gives me the exact same answer. Only instead of using the OpenAI playground to submit the question, I write code to make a network request to the OpenAI web API.
Can I expect the question/answer to the web API to cost the same, in tokens, as my OpenAI playground questions/answer? My educated guess is ‘yes,’ that tokens calculation doesn’t vary based on how the question is submitted. But just want to confirm here.
Can I expect the question/answer to the web API to cost the same, in dollars, as my OpenAI playground questions/answer? If not, how might it vary?
Context: As a side project I want to build a web app which will ask ChatGPT many questions via the web API. Right now I want to estimate how much those requests will cost me, financially.
You can get a Plus subscription for priority access and access to new beta features as they are released, early. This is charged at $20/mo + local VAT tax rates.
The OpenAI GPT API endpoints are priced at various levels for prompt and completions (input/output) details of which can be found here Pricing
I don’t think ChatGPT does only utilize just one model e.g. GPT-4 (is GPT-4 even just one model?).
And if I understand it correctly you want to scrape ChatGPT?
Playground is just a UI over the API, if you look at network traffic it’s calling the same API endpoint you would use in your projects. So yes Playground and API should cost the same.
Yes, the playground requests and requests you issue yourself using curl, python, node, or any other API client, work exactly the same.
The playground is just a graphical interface on top of the API.
Playground is a quick and easy way to explore the OpenAI API, the pricing is going to be the same as making API calls from your own code - unless the playground is doing additional processing.
unless the playground is doing additional processing.
Can you give me an example of what you mean by this?
Can you expand on this? As I understand it, every request sent via chat.openai.com or the OpenAI playground requires users to first create an OpenAI account, then after, requires tokens. New OpenAI accounts are given X tokens to use for free (for a limited time), but after those tokens are expended (or expire), new tokens come at a monetary cost. What am I misunderstanding?
