How to create an OpenAI deployment with adequate resources?

Question

Hi,

I cannot figure out how to deploy and OpenAI model with enough resources for us to actually use it outside of testing. We having been using the sponsorship subscriptions and that has been helpful for testing/development, although we do regularly hit the 100 RPM limit. Now I need to deploy something for beta testing and I cannot have the 100 RPM limit. I tried creating a paid subscription and deploying a model against it, but the limits were even worse than the sponsorship limits. The support portal told me to use the form to request an increase, but that form does not load it just says that the form is closed. I'm not sure what to do at this point, I'm about to look at alternatives. The rate limits if I go directly to OpenAI are 10K RPM and 30M TPM for gpt-4o. The limit I was getting through Azure was 30K TPM. It's just not usable. How do I actually use Azure OpenAI for a real application?

Thanks,

Brian

Answer 1

Hi ,

Thanks for reaching out to Microsoft Q&A.

It can be frustrating that the request form is frequently closed and that default quotas are too low for production scenarios. However, most Azure OpenAI customers who can demonstrate real production needs eventually get a quota increase, often through a direct conversation with Microsoft Support or their account manager. Persistence and clarity about your business case are key.

If none of these work (or if your timeline is more pressing than microsoft’s typical approval process), using OpenAI’s own platform (api.openai.com) with its more generous default caps might be a more viable short-term solution. Then, once your Azure OpenAI quota is approved, you can integrate back into Azure for improved security, compliance features, or private connectivity.

Hopefully this clarifies how to navigate azure’s sometimes tricky quota landscape, and you can get the throughput you need for a real production application.

Please feel free to click the 'Upvote' (Thumbs-up) button and 'Accept as Answer'. This helps the community by allowing others with similar queries to easily find the solution.