I guess I am not “eligible” yet since I don’t see it in the base model dropdown. But how do I become eligible? How do I request access to fine-tuning for GPT-4o?
Start fine-tuning 
Once you have a bit of a track record in fine-tuning, you normally become eligible (to request access).
1 Like
So I need to do some fine tuning for, let’s say, gpt-3.5-turbo, even though I don’t intend to use gpt-3.5-turbo? How much ‘track record’ do I need?
(for obvious reasons, I would prefer to not fine-tune something I won’t be using)
Here’s a link. If you don’t have a reasonable presentation of why your application failed with fine-tuning gpt-3.5-turbo, and why you will provide even more resources to gpt-4 and expect success that could not be accomplished otherwise (and are a company willing to fund development of thousands of examples), you’ll likely never hear back.
2 Likes
Well, I think the intention behind this is to first establish whether you could fulfil your fine-tuning needs with a gpt-3.5 model and, more generally, evaluate whether fine-tuning is the best route for your use case.
You can give it a try to directly navigate to the form but I am just not sure if you will even be considered then without any prior fine-tuning experience. But if you have a compelling use case, you might be lucky.
Well, the reason we want gpt4o is because we do large document analysis. The documents regularly exceed the context window of gpt3.5. Not sure if that is a good enough reason for openai.
Regardless, thank you for sharing the link. I can only try.