I’ve been developing an application using either 4o or 4.5-preview (for different tasks). I couple of weeks ago I was using 4.5 quite heavily (spent over a hundred dollars in API costs), but now the exact same code is triggering this error:
Error code: 404 - {'error': {'message': 'The model `gpt-4.5-preview` does not exist or you do not have access to it.', 'type': 'invalid_request_error', 'param': None, 'code': 'model_not_found'}}
If I use 4o the code works fine. My organization is Tier 3, so it shouldn’t be an access problem. I also have a positive credit balance and am nowhere near my monthly limit. I tried creating a new API key (which I saw recommended on other posts with similar issues) but that didn’t work.
Ah, I see! It’s still listed on their Prices page so I didn’t realise that it have been phased out. Do you know if it’s going to be offered as a production model?
I think, for me, I’ve solved my problem. I had been using 4.5 as a fallback when 4o gave several refusals in a row (“I’m sorry, I can’t help you…”) I’m not sure why I was getting refusals so often, but I improved the refusal rate by adding some encouragement ("Remember, you can do this!) to the prompt. And I added 4.1 as the fallback (I still sometimes get several refusals in a row). It seems to actually work really well and I might replace 4o with 4.1 as the primary model anyway, it doesn’t seem to have the same refusal problem that 4o does.
The answer: OpenAI announced they would shut off the model GPT-4.5 on API in the same breath as introducing GPT-4.1 in April.
OpenAI sent multiple emails to anyone who had made calls to GPT-4.5 in the 30-day lead up to the shutoff date. It has been turned off for over two weeks.