I asked support what I could do with the money I accidentally paid for OpenAI’s API service. They answered, “Access OpenAI Models: You can use the credits to access various OpenAI models, including GPT-4o, GPT-4o-mini, DALL E, Whisper, and more. These models can help you with tasks like text generation, image creation, speech recognition, and more.” But what’s the difference between using a Plus subscription or using this OpenAI’s API service? The Plus subscription already has everything that support wrote about.
There are some AI services online that require you to add an OpenAI Key to use them.
But OpenAI does not recommend that.
So what I could imagine is you could cancle the plus subscription, then install “docker” on your computer, open a terminal and use this command:
docker run -d -p 3000:8080 -e OPENAI_API_KEY=your_secret_key -v open-webui:/app/backend/data --name open-webui --restart always ghcr.io/open-webui/open-webui:main
You need to login to https://platform.openai.com and create an OpenAI Key and replace the your_secret_key part of that command with your key - which should start with sk…
it will then look like this in your terminal:
And when that is completed
then open http://localhost:3000 in your browser (give it a a minute or two to start up before you open the browser).
It is like your own private ChatGPT locally which uses your API money.
You register your admin account.
and when logged in you can see all the models you have access to and you can select one and start your chatting…
You can even export ChatGPT data and import it there.
Might even be possible that when you don’t use ChatGPT a lot that it is a cheaper option over time.
https://platform.openai.com/usage << this is where you can check how much that costs when you chat with it. Like run a Chat, then go into the usage overview and you will see…
Ah and here you see how much it costs to use the different models over the API:
4 Likes
Thank you, jochenschultz. Can I ask a couple more questions? I have already paid for the subscription plus for a month. If I don’t use the balance within the month, how long will the $20 remain on my account?"
12 months is written in their terms…
1 Like
API credits purchased by mistake or with regret should be able to be refunded under the contract laws of many countries for a period, despite what OpenAI says. If you would never find API use, you directly would request a refund for unintended purchase. Your initial date expressing the desire fixes the need for compliance.
Uses? Use models not available in ChatGPT, such as GPT-4 (real original, not what ChatGPT brands as gpt-4), o1-preview (again).
Language models can be used in the platform site with “playground”->“chat”.
Use embeddings for AI similarity of texts and documents. Voice transcriptions and speech. Documentation linked on the right will give an idea of the API endpoints.
Thank you. Are there any differences between using ChatGPT Plus directly and accessing the API, as recommended when working with Docker? What are the potential limitations and benefits?
Thanks for the clarification
There are some but it really requires to know what kind of stuff you want to do.
And I don’t think in the first 12 month of using it that would be relevant at all.
I frequently use ChatGPT for advice on Java, Android, and Python code, improving my documentation, reports, and English phrasing. I also regularly ask about history and literature.
The main difference between ChatGPT and the OpenAI API is as follows:
ChatGPT is aimed towards non-programmers, helping with everyday situations and a finished application; a chat app.
The OpenAI API is aimed towards programmers, it lets you utilize the power of AI within your own application.
So you are not only limited to making a chat app, you could make a story generator, a PDF validator or anything you need creativity or some intelligence for.
Basically:
- ChatGPT is an app to “play around” with AI
- The API is meant for developers to integrate into their applications to control AI programatically
Cheers. 
Thanks. I’m familiar with the concept of public APIs. Can I create Java client to use this API?
There’s quite a few clients and libraries available.
Do you have a need for images? That’s a nice way to go through the tokens and have something of value at the end.