Is RFT actually available to the public? If not, when?

API
1

I have browsed the internet and various documentations (and some cookbook), and I have tried implementing a simple RFT myself. But I keep getting this error:

openai.BadRequestError: Error code: 400 - {‘error’: {‘message’: ‘Invalid model o4-mini for method reinforcement’, ‘type’: ‘invalid_request_error’, ‘param’: None, ‘code’: ‘model_not_allowed’}}

So, am I right that it is still not open to the public or what is up (sorry if it is an error on my end)? And if it is indeed not available for everyone, how to apply to join the set of organisations who can try it out and potentially use it? (in my case, for educational purposes, trying to make a better grader actually).

2

Reinforcement fine-tune of o4-mini should be an example of gatekeeping behind the ID verification.

image

“subject to org verification”: “blocked unless you submit to personal government ID scans and having your selfie video sent to a sketch 3rd party company.”

Check the fine-tuning “create” in the platform site dashboard to see if it is offered to you there:

image
image587×319 5.41 KB

1 Like