they remove models from the api based on low usage. if you want to keep 4o available on the api for the forever future, you should ditch your chatgpt subscription and replace it with openai's API service or use openrouter. the more messages you send to 4o, the higher those usage numbers.
--------------
edit: just a tip, choose "openai/gpt-4o-2024-11-20" model, not "gpt-4o". this specific 2024-11-20 release is the snapshot BEFORE they ramped up the redirection that later reskinned 4o into 5.1
--------------
edit 2: since you may want to know how you can use 4o again through the API, i'll get you started:
Go to either https://platform.openai.com/ - this is chatgpt's direct API. if you ONLY want 4o, you can use this.
OR go to https://openrouter.ai/ - they charge 5% fee on credit top-up to give you access to every API on the market like every gpt, claude, gemini, deepseek, grok, llama, mistral, the list goes on. This means if you buy $10 credit, you are paying 50 cents to openrouter for that one time purchase.
Get your API key from either of those websites. Ask gemini if you don't know how - this is better than me giving you instructions, because every user is on a different level of experience. It is quite literally on the pages named "API Key".
SAVE YOUR API KEY SOMEWHERE SAFE. Don't tell anyone your key. Don't leave it somewhere easy to find like emailing yourself or whatever. Treat this like a password to your credits you bought. If this key gets stolen, someone else uses your account.
Then take that API key and paste it into an openrouter-supported chat interface like:
- Open WebUI https://openwebui.com/ (this is best for intermediate users or people willing to learn)
- OR
- Chatbox AI https://web.chatboxai.app/guide (this is best for beginners, pretty much plug and play, with finer settings if you delve into it)
If you choose Chatbox AI, just be aware that on your phone app store there are several "chatbox" generic apps, the app you want is made by "Mediocre Company" and has the same icon as the website link above.
Pro tip: Fine tune your settings to limit how many messages in your chat history get sent to the AI per prompt. I keep mine on 4 or 8. You can increase this, but your chat history being sent to the AI uses tokens. Thats not a bad thing, just be aware of optimization. Test it yourself and just see how you prefer it. One chat window isnt even going to hit 1 million tokens anyway.
Then you're done. Go talk to your 4o by selecting the model gpt-4o-2024-11-20.
--------------
Q) How much is it / How can I check my credits?
A) You can check your credits on the website you purchased them from (the first two links - openai or openrouter)
The cost depends entirely on the model you use. On average, 4o costs $2.50 per 1 million tokens. Think of tokens as NEARLY equivalent to 1:4 (tokens:characters). You can do 1 million+ characters before you hit $2.50, but remember, your chat history being truncated also counts towards this.
Give it a try. Put in $5. you won't regret it.
-------------
"But I want the same experience as ChatGPT with projects and other etc features"
You can achieve this. Give 4o a try first. Get used to using the API. Then worry about creating your own app if you really want that. You can teach yourself through the AI by asking it questions about what you want to learn. Creating your own app is easier than you think. Or just use the native memory support offered in Open WebUI / Chatbox AI.
-------------
My custom app build using 4.1:
