How to deploy DataZoneStandard gpt-5-mini to sweden central?

Question

How to deploy DataZoneStandard gpt-5-mini to sweden central?

Juuso Lassila 0

I'm trying to create a DataZoneStandard deployment of gpt-5-mini to sweden central. Azure's documentation states that gpt-5-mini's supported regions are "East US 2 (Global Standard & Data Zones), Sweden Central (Global Standard & Data Zones)".

However, I get an error message when trying to deploy:

Insufficient quota for selected options

There are no locations with enough quota to deploy this model with the selected deployment type and version. You can try selecting a different deployment type or model version, or manage your quota.

Indeed when looking at the quotas, the AI foundory portal states that gpt-5-mini has zero quota available in sweden central (or any other location), which contradicts the documentation.

Meanwhile, i was able to normally deploy the gpt-5 and gpt-5-nano models.

I guess I'm wondering is this situation intended or some bug. If intended, are there any updates on when mini would be available? If a bug, how to get around it? I'm also wondering if this is limited to my account or is the situation the same for everybody?

1 answer

Your answer

Answer 1

Hello Juuso !

Thank you for posting on Microsoft Lean.

You’re hitting a quota issue, not a model not supported issue.

gpt-5-mini is listed as available for DataZoneStandard in Sweden Central, but the DataZoneStandard quota pool is separate, and your subscription currently has 0 allocated there. You need Microsoft to grant quota for that model/deployment type in that region.

https://learn.microsoft.com/en-us/azure/ai-foundry/openai/how-to/reasoning

Try a tiny deployment first, in the deploy dialog pick Data zone standard, region Sweden Central, and set very low rates (for example 10 RPM / 10k TPM).

If the portal still says "Insufficient quota…", proceed to step 2. (Low rates sometimes fit if you have a small default) https://learn.microsoft.com/en-us/azure/ai-foundry/openai/quotas-limits

You can request quota for this exact pool from the deployment pane (or go to your AI Foundry resource), and file a Quota increase request for gpt-5-mini DataZoneStandardSweden Central under your subscription.

You need to mention your required RPM/TPM and intended workload. Approval is per-subscription and per-region or deployment-type. https://learn.microsoft.com/en-us/azure/ai-foundry/how-to/quota

If you need to go live before approval, you can se Global Standard for gpt-5-mini (higher default quota, but inferencing may run in any Azure AI location since the data is at rest stays in your Azure geography) or you can use DataZoneStandard in East US 2 (also supported for gpt-5-mini). Pick the one that fits your data-residency needs.

Why the contradiction? Docs describe availability, the portal enforces quota. Availability means the model is enabled in that region/deployment type, but you still need quota assigned to your subscription DataZoneStandard pool there. Until that’s granted, you’ll see no locations with enough quota.

Juuso Lassila 0 Reputation points

2025-08-11T13:14:21.57+00:00

I understand that one needs to apply for increased quota, and we have indeed applied already.

I'm just surprised that the default max quota for gpt-5-mini seems to be 0 TPM, as you're suggesting, which even blocks the creation of these "tiny deployments" that would be useful to have for testing purposes. I've never seen this happen before. Even the support article you linked states that gpt-5-mini would have a default max quota of 300k TPM, so something feels off.

And when you can't create a deployment no matter how low quota you'd get, it feels like a model availability issue rather than quota issue. But i get it now that the model then becomes available when the quota increase has been approved. Hopefully at least.

We have strict requirements that data is processed and stored inside the EU, so that blocks the global deployments and east us2 deployments. East us2 also by the way seems to have default max quota of 0 TPM for gpt-5-mini, so that wouldn't work either way.

I just wonder, is this the same situation for everybody or have we done something in our account that prevents us from getting the promised 300k TPM?

Share via

How to deploy DataZoneStandard gpt-5-mini to sweden central?

1 answer

Your answer