Foundry quota documentation: concern about quota increase.

[shodo]

Hi everyone,

In my company, we are managing quota allocation for different deployments. Currently, we have multiple deployments of a GPT-4.1 model with a shared quota of 9M TPM total. This means, for example, that I can have 9 deployments, each allocated 1M TPM.

The point is that documentation here states:
_

Use the shared quota only for creating temporary test endpoints, not production endpoints. For endpoints in production, you should request dedicated quota. Billing for shared quota is usage-based.

_

So seems that even 9M TPM are sufficient, It's a best practice to request "additional" tokens reserved for production endpoints.
But, at the same time, when opening the module to request additional quota, I read:

Priority will be given to customers who generate traffic that consumes the existing quota allocation, and your request may be denied if this condition is not met.

It seems contradictory: the doc says I need to request additional quota because the shared one is reserved for testing purposes. However, my quota increase is denied if I don't use all the available quota (which I shouldn't use at all for production purposes!).

There are also two different doc pages:
- Manage and increase quotas for resources with Microsoft Foundry (Foundry projects)
- Manage Azure OpenAI in Microsoft Foundry Models quota
I'm having trouble understanding the difference between Azure OpenAI managed models and the overall models available in Foundry in terms of quota. It seems like Azure OpenAI is just a subset of the models offered in Foundry, so the first page should be valid for the second one. But, the Azure OpenAI page doesn't mention how to request a quota increase for production use. How do the two fit together, and which is the best practice regarding shared quota usage?

[Jothsna Praveena (jothsnapraveena)]

I think the confusion comes from the term “shared quota” being used in two different contexts.

If the 9M TPM is the GPT-4.1 quota allocated to your Azure subscription, you can distribute that quota across multiple deployments and use those deployments in production. You don't need to request an additional 9M TPM simply because the endpoints are production.

The “shared quota is for temporary/test endpoints” guidance refers to Foundry’s temporary shared quota pool, which provides capacity when your subscription doesn't have sufficient dedicated quota. That shared pool shouldn't be relied on for production.

The statement about quota increases also makes more sense with this distinction. If you already have 9M TPM and request additional quota, Microsoft may prioritize the request based on how much of your existing 9M you're actually consuming. They aren't saying you need to consume the temporary testing quota first.

The two documentation pages have different scopes as well. The Foundry quota page covers the broader Foundry model ecosystem and shared quota mechanism, while the Azure OpenAI quota page covers quota allocation for Azure OpenAI models such as GPT-4.1.

So in this case, if the 9M TPM is your subscription's GPT-4.1 quota, using it for production is fine. Request more quota when your existing allocation is becoming insufficient, not simply because the workload is production.