Can people purchase only the API Access alone, and not have to pay for the “Plus” subscription? The docs don’t make this clear and only says they’re billed separately.
1 Like
They are entirely different products.
You may choose to use none, either, or both.
3 Likes
Thanks! I just purchased $5 worth of API usage to try out. Looks like you can pay in advance whatever amount you want.
Whatever you want, as long as you want a minimum payment of $5 and a maximum payment and balance of $50.
The prepay system is new, made to work in OpenAI’s favor, with kinks to be worked out. API users previously would be billed at the end of the month.
1 Like
So can I use the GPT-4 API witout the ChatGPT plus subscription? Or can I only use GPT-3.5 if I don’t have ChatGPT plus?
ChatGPT Plus and the API are completely separate billing. If you want to use GPT4 or plugins on chat.openai.com you need Plus.
Plus does not get you GPT4 on API.
1 Like
If you only need access to the API you can pay for that alone. That’s what I did. The API will give you access to both gpt-3.5 variations and gpt-4 (via the API of course, not GUI). Since the only way I consume GPT-4 is thru my own app GUI I therefore never had to purchase “GPT plus”.
1 Like
Great, this is what I’m looking for, I purchased a subscription to ChatGpt, and now I am looking to start using Obsidian with a plunging, so I’ll try both paid-for API tokens and GhatGPT and then just keep one; my thoughts are to keep all my chat organized and nothing better than Obsidian.
1 Like
How did it go? I’m on the same boat right now but just wondering if I should cancel the plus subscription. If it’s too much more expensive as a heavy user I may have to keep plus
ChatGPT: costs you fixed money regardless of whether you use it
OpenAI API: costs you nothing if you don’t use it, increasing amounts the more you use it.
With API, you’ll need to develop a software platform where you have the same features of Plus that you actually use. With API, you are not limited to just an AI that chats with a user in a conversational manner.
The decision is really the anxiety you feel about running out of GPT-4 turns in three hours on Plus, vs the anxiety of every input and every growing conversation costing you just a little bit more than $0.00.
While creating a GUI might be a hassle, if you’re just looking to chat, you don’t need a complex GUI. You can find simple, self-hostable GUIs on GitHub.
For vision-language tasks and other complex language challenges, you’ll need a commensurate level of expertise.
By the way, whether it’s GPT-3.5 or GPT-4, in English, one word counts as one token, but in Japanese, one character counts as one token or more, which can subtly impact the overall efficiency and indirectly affect the operational costs.
1 Like