Will existing APIs be automatically upgraded when GPT-5 comes out?

Of course it's not automatic, right? You have to manually change the model name lol

There’s no such thing as an automatic upgrade. Back when GPT-4 came out, existing 3.5 users just had to stay where they were—if you wanted to connect to the new one, you had to manually change the endpoint or model ID. Fortunately, if you explicitly specify “gpt-4o” now, that model stays as-is; it won’t suddenly switch just because GPT-5 is released. That said, if OpenAI marks it as deprecated, you’ll need to migrate at that point. Checking the official docs is the fastest way.

Well, 'auto-upgrade' might actually be a risky proposition. When the model changes, the nature of the output changes too, so swapping it in automatically would obviously break the system. So even if it were done automatically, I'd say it's something you need to manually prevent.

Short answer: no. However, if you don't specify the model name and just use the endpoint, the default version could change—that's risky, so always hardcode the model name. When GPT-5 comes out, OpenAI will provide guidance on that, so just follow it. Migration notes will follow after that too.

Oh, I was curious about this too. But hasn't GPT-5 even been officially announced yet? Feels a bit early, lol.

I don't know much about APIs, but from what I understand, whenever a new model comes out, everyone integrates it fresh. It probably isn't automatic.