Team Subscription - No access to gpt-4-32k API

Hi,

With the introduction of OpenAI teams, OpenAI explicitly said the subscription would get access to the 32k context length model of gpt4:

We have a running Team subscription on our account, but we still can’t access the “model”: “gpt-4-32k”, via API, with the response “The model gpt-4-32k does not exist or you do not have access to it.”

So now I’m wondering if we misunderstood, and the 32k context length model is not available via API after all? We dont have it in the playground either.

Thanks in advance! :smile:

1 Like

Hi!

The API is still a separate service from Team.
I’d assume you have to purchase 5$ in credits and then you will get access.
You can check if my assumption is correct by calling the gpt-4 model. If that also doesn’t work then you may have to buy credits.
If you have access to gpt-4 but not the 32k variant I suggest you contact support.

Hope this helps!

1 Like

Thank you for the reply. I have access to the regular gpt-4 model on the API and have been using it for a while, so maybe I’ll have to contact OpenAI and ask. (I have credits available on the account).

1 Like

Ok, hope this will get sorted out soon!
By the way, can you see the 32k model in the playground when using the selector for the chat models?

Unfortunately no, I didn’t get any new ones after changing plan.

Largest context length I’m able to get is 8k

May I ask why you can’t use one of the gpt-4-turbo models instead? It has a much larger context window than even the 32k model and it’s more than 4x cheaper.

As shown in the linked post, the 32k model was a very limited rollout and probably won’t become widely accessible since gpt-4-turbo is faster, cheaper, and supposedly better.

1 Like

The gpt-4-turbo-preview (gpt-4-0125-preview, gpt-4-1106-preview) are 128K context models, which are almost certainly the same language models used by the ChatGPT teams.

It is highly likely that the ChatGPT teams use a shortened version of the context from the aforementioned models.

The original 32K context language model is hardly used nowadays.
It is believed that the turbo model will be the mainstream model in the future.

1 Like

Still, if it’s in the service description then it should be available.
Or the description has to be updated.

1 Like

Reading their plan comparison page it seems that the 32k context model is within ChatGPT, not the API.

I do agree though that the wording in the screenshot from @cisnes leaves far too much room for confusion.

2 Likes

Thank you!
This sounds like the solution to this question from OP.

1 Like

How confusing!
Anyone who sees this screen would think they can use it with an API.

But be careful, we’re dealing with OpenAI here!

Could you check the account you are accessing it from?? You could be logged in using the Email ID added to Teams and still be mapped to your personal account. You need to switch the account to gain access to GPT-4.

You can do that by clicking on the “A” Icon on the navigation menu ( at the bottom-left corner) and selecting the right account.