Using newest model instead of depercated

So far I have used the model text-davinci-003 for my experimental application. Now I would like to test gpt-4 (turbo?). I’m not a real developer (I use some no-code tools like webflow and wized) and wanted to replace “text-davinci-003” with “gpt-4-1106-preview”, but that doesn’t seem to work. Can anyone help me?


1 Like

You’ll need to complete at least one successful API payment before you’ll be granted access to the newer models. You can pre-pay; I put $5 in and was granted GPT-4 access in a few minutes after the payment processed; and GPT-4-1106-preview shortly thereafter (when it was released).

Go to - then in the left sidebar, click Settings, then Billing, and there will be a button labeled ‘Add to credit balance’ that will let you fund your API account. That should do the trick.


@idotter it may be possible that they have not given you access to “gpt-4-1106-preview” yet.

One way to verify this is to see if you have access to that same model in the playground here:

Try to see if you can select gpt-4-1106-preview there. If it’s not available it means you havent been given access yet

and if not try what @erikafoxelot suggested!

1 Like

text-davinci-003 is a completion model that takes a “prompt” and is used on the completion endpoint.

A replacement model that would drop in on the completion endpoint is gpt-3.5-turbo-instruct, except that it will be different in its responses, of course.

To use a chat model, you will need to use the chatcompletions endpoint, use formatted json list of role messages, and obtain the results from a different return object.

API reference is on the sidebar of the forum.

THX for all your replys … seems to be the stuff with /completion and /chat models … @_j is there somewhere a description when to use which model? :innocent: