Can GPTs use GPT4 instead of GPT4o?

Honestly GPT4o seems less good then GPT4 in many cases, it seems less conversational in outputs and always gives such long lists. I’ve also found it to be less accurate and have tested a few times where GPT4o gave me the wrong answer and I switched to the older one and got the right answer.

1 Like

It’s a little bit finger in the wind, but I wanted to add that I agree with the above. For instance, when asking coding questions and then following up with additional info/questions, instead of just writing the bits of the code that have changed (like <=4 did), instead 4o will rewrite the whole thing with maybe only one line changed, even when I explicitly ask it for the changed lines only sometimes. Obviously it must be such a difficult product to regression test, but I’ve always found the models felt like they got significantly better until the 4 → 4o update. Obviously the speed increase and cost decrease makes it good for a lot of things though, don’t get me wrong!

I completely agree that GPT-4 is better than GPT-4o. I don’t think OpenAI has the incentives to switch it back to GPT-4 though. It is cheaper to run GPT-4o for them due to its design.