Azure OpenAI Service Launches GPT-4 Turbo and GPT-3.5-Turbo-1106 Models

keijik

Microsoft

Nov 27, 2023

HI again - I hope I have answered all your questions & comments here:

"Current token count" only supports up to 8192 in Studio Chat UI

We are aware of this issue and currently working on fixing it. API supports 128,000 tokens

Model accuracy comparison with prior versions

We are interested in details of reports of model quality regressions. You can link to details here, or If you want to send us your results privately, please contact me and we can investigate: https://www.linkedin.com/in/keijikanazawa or https://twiter.com/gojira (DM).

We are aware that in certain cases the output is qualitatively the same from human perspective but might be shorter / longer or formatted differently.

Pricing

The pricing is as described above, they should be published on azure.microsoft.com by end of week.

Model Not Showing in Studio

Model availability by region is kept up to date here: https://aka.ms/oai/models
In Studio you can first check in the "Models" tab for GPT-4 1106-preview
Assuming you see GPT-4 1106-preview in the Models tab, then for deploying, select "gpt-4" and then be sure to select "1106-preview" in the dropdown (even if it says you have no quota, because quota for GPT-4 Turbo is separate)

Resource Not Found 404 with API call

This may be due to the endpoint URL or possibly due to API version used in API call

When available in XYZ region?

We have no updates to share at this time - this page is kept up to date: : https://aka.ms/oai/models

No region with GPT-4 Turbo, DALL-E 3, and Whisper

Thank you for your support of Azure OpenAI Service! We have no updates to share at this time but we will look to enable this in future.

Blog Post

Azure OpenAI Service Launches GPT-4 Turbo and GPT-3.5-Turbo-1106 Models