Azure OpenAI instance, modifying deployment capacity to permit another model deployment

TH-4622 20

I have an Azure OpenAI instance with multiple models. I want to deploy another model, so modified the deployment capacity of one of them (4o) downwards from 1000K TPM to 950K TPM to permit another 4o model deployment within quota. I selected Standard deployment type. This didn't allow me to deploy the model... it states "Sweden Central (no quota)", the only location that it will permit me to deploy to is Japan East, even when the TPM limit I selected was very low, e.g. 20K TPM.

In AI Foundry, under Quota for this 4o model it states Quota Allocation: 950K of 1M, which suggests to me there should be capacity? Maybe there's also a limit of number of models per region? Seems odd that I can deploy to Japan East but nowhere else? For data sovereignty reasons within the firm I am only permitted to deploy to EU, hence Sweden Central.

What do I need to do to be able to deploy the additional 4o model please?

Saideep Anchuri 3,140 Reputation points Microsoft Vendor

2025-02-27T13:37:48.1933333+00:00

Hi TH-4622

Sorry for the trouble. You can either TPM to lower than 700/800k TPM for first model deployment or change deployment standard to global standard or Data zone standard to overcome this issue. Global standard deployment in any region provides much higher quota which dynamically routes the request based on best availability.

Kindly refer below link: deployment-types

Thank You.
TH-4622 20 Reputation points

2025-02-27T16:37:27.02+00:00

Thanks Saideep, I have reduced down to below 700k TPM, which leaves over 300k TPM available of the quota on 4o (and left it for several hours to apply) but it still only permits Japan East as a deployment location. As stated above I am only permitted by our internal firm policies to deploy to the EU, so am still stuck on this.

Accepted answer

Saideep Anchuri 3,140 Reputation points Microsoft Vendor

2025-02-27T17:04:32.0966667+00:00

Hi TH-4622

You, must be going with default version 2024-11-20 which selects Japan east as default region , please change to model version to 2024-05-13/ 2024-08-06 to create model deployment with 600k, 300k.

Kindly refer below screenshot:

kindly refer below link: deployment-model-availability

Thank You.
Please sign in to rate this answer.

1 person found this answer helpful.
Saideep Anchuri 3,140 Reputation points Microsoft Vendor

2025-02-28T06:00:19.5+00:00

Hi TH-4622

Following up to see if the above response was helpful.

Thank You.

TH-4622 20 Reputation points

2025-02-28T08:10:29.75+00:00

Hi Saideep,

Thanks, changing to an earlier date than the latest in your recent answer does allow me to create the deployment... but there was no drop down for location for them, so I assumed this meant that it was the default Global option which is not permitted by our firm. I tried it anyway and can see that it has created in the same region as the previous model (Sweden Central). This was not made obvious in the interface. I'm unsure of what relevance there is of using an earlier date though. I had assumed that that latest date must be the best. Perhaps this is not the case?

Saideep Anchuri 3,140 Reputation points Microsoft Vendor

2025-02-28T08:59:32.6433333+00:00

Hi TH-4622

We have provided the suggestion based on model summary documentation. There is no mention of 2024-11-20 version for Standard deployment table.

Though 20-11-2024 is coming in dropdown, but it shows only Japan east for supported, region.

Hope you understand. Please upvote the answer if it helped address your issue.

Thank you.

TH-4622 20 Reputation points

2025-02-28T10:48:31.0533333+00:00

I think MS UX team need to work on their interface. Thanks Saideep. I have marked your answer and upvoted.
Sign in to comment

Use comments to ask for clarification, additional information, or improvements to the question.

Share via

Azure OpenAI instance, modifying deployment capacity to permit another model deployment

0 additional answers

Your answer