Azure AI Foundry – Research Models Auto-Deploying to Sweden Central and Hanging on Quota Load After Hub Deletion

Question

Azure AI Foundry – Research Models Auto-Deploying to Sweden Central and Hanging on Quota Load After Hub Deletion

Red 25

I was previously approved to use models like gpt-5-codex and sora-2, and everything worked fine through the Responses API.

After reading that OpenAI’s Codex could be accessed through Azure as a provider, I attempted to configure that setup but couldn’t get it to work. I then discovered that I could use the Azure AI Foundry extension for VS Code to run gpt-5-codex through GitHub Copilot, so I set that up next.

During configuration, I ran into a hub-related error. Thinking it might be easier to start fresh, I deleted my existing hub and created a new resource group and hub. I suspect this is where things started to break.

Now, whenever I try to deploy any research models (released models still work fine), two things happen:

A new Foundry is automatically created, and

The model is deployed to Sweden Central, it shows no quota elsewhere.

This feels like it might be a quota or resource-linking issue, but here’s the odd part: If I go into the Chat Playground on useast2, click Create New Deployment, and select gpt-5-pro, the interface hangs indefinitely on “Loading resources and quota”.

From my understanding, everything should be tied to the subscription ID, so deleting the hub shouldn’t have broken the linkage, yet it seems something important seems to have got disconnected.

Has anyone run into this before or know what the correct recovery steps might be to get the research model deployments working again?

Anshika Varshney 1,910 Reputation points Microsoft External Staff Moderator

2025-10-23T20:09:19.21+00:00
Hello Red,

As per your latest comment, I can see that your quota page still shows models under East US 2, even though you’re expecting them to auto-deploy elsewhere. This is expected behavior because:

Quota is region-specific: Even if Azure AI Foundry auto-deploys models globally, the quota view will still list the region where the resource is provisioned (e.g., East US 2).

GlobalStandard tier means the model can be accessed from multiple regions, but the quota management UI reflects the original region for tracking purposes.

Auto-deployment does not change the quota region it only ensures the model endpoint is available without manual deployment steps.

What to check next

Confirm that your endpoint URL works and is serving requests (regardless of quota region).

If you need quota in another region, click Request quota for that region in the portal.

For models like gpt-4, gpt-4o, or gpt-4o-mini, auto-deployment is handled by Microsoft; you don’t need to manually deploy them unless you want a dedicated capacity.

In short: The quota page showing East US 2 is normal and does not block usage. If you need higher limits or a different region, request quota for that region.

Reference: Manage Azure OpenAI in Azure AI Foundry Models quota

I hope this has been helpful! Let me know if there are any remaining questions or if further guidance is needed. I’ll be happy to assist.
Red 25 Reputation points

2025-10-23T20:39:00.7666667+00:00
This was not my issue. The target uri and endpoints for each model deployed on swedencentral were not populating at all. What is expected is a warning of no quota, as you suggested. However, a new resource was being auto deployed when inside Azure AI Foundry -> Hub -> Project using a specific region + deploying a model.

Issue was solved by submitting a support ticket and having them resync with the new foundry/hub/project.

For your awareness, this was my flow inside of a project (this is with the issue solved so I will add notes):

Here before the fix "Resource Location" would be empty

A new resource would be created with it's own set of models.

This should not happen:

Issue is solved! Thanks for your help.

Answer accepted by question author

0 additional answers

Your answer

Anshika Varshney 1,910 Reputation points Microsoft External Staff Moderator

2025-10-23T20:09:19.21+00:00

Hello Red,

As per your latest comment, I can see that your quota page still shows models under East US 2, even though you’re expecting them to auto-deploy elsewhere. This is expected behavior because:

Quota is region-specific: Even if Azure AI Foundry auto-deploys models globally, the quota view will still list the region where the resource is provisioned (e.g., East US 2).

GlobalStandard tier means the model can be accessed from multiple regions, but the quota management UI reflects the original region for tracking purposes.

Auto-deployment does not change the quota region it only ensures the model endpoint is available without manual deployment steps.

What to check next

Confirm that your endpoint URL works and is serving requests (regardless of quota region).

If you need quota in another region, click Request quota for that region in the portal.

For models like gpt-4, gpt-4o, or gpt-4o-mini, auto-deployment is handled by Microsoft; you don’t need to manually deploy them unless you want a dedicated capacity.

In short: The quota page showing East US 2 is normal and does not block usage. If you need higher limits or a different region, request quota for that region.

Reference: Manage Azure OpenAI in Azure AI Foundry Models quota

I hope this has been helpful! Let me know if there are any remaining questions or if further guidance is needed. I’ll be happy to assist.
Red 25 Reputation points

2025-10-23T20:39:00.7666667+00:00

This was not my issue. The target uri and endpoints for each model deployed on swedencentral were not populating at all. What is expected is a warning of no quota, as you suggested. However, a new resource was being auto deployed when inside Azure AI Foundry -> Hub -> Project using a specific region + deploying a model.

Issue was solved by submitting a support ticket and having them resync with the new foundry/hub/project.

For your awareness, this was my flow inside of a project (this is with the issue solved so I will add notes):

Here before the fix "Resource Location" would be empty

A new resource would be created with it's own set of models.

This should not happen:

Issue is solved! Thanks for your help.

Answer 1

Azar 30,735 MVP Volunteer Moderator

Hi there Red

Thanks for using QandA plaform

the hub maintains the linkage between your subscription and the research models. Once that hub is removed, Foundry can no longer resolve where the research quota applies, so it defaults to a fallback region (like Sweden Central) and stalls when attempting to load resources.

Try this, verify under Azure AI Foundry → Manage quotas whether your research models still appear. If they don’t, you’ll need to open a support ticket with Microsoft and request that your research quota be remapped to the new hub or resource group.

If this helps kindly accept the answr thanks much.

Red 25 Reputation points

2025-10-23T18:51:00.8166667+00:00

This is what I see when I look at manage quota. They still appear with 'East US 2'
Red 25 Reputation points

2025-10-23T20:21:26.04+00:00

Created a support ticket for this issue for them to resync. Issue is fixed - this comment helped me realize that the models were still showing as deployed (hence the quota limit).

Share via

Azure AI Foundry – Research Models Auto-Deploying to Sweden Central and Hanging on Quota Load After Hub Deletion

0 additional answers

Your answer