Microsoft Foundry: I can not deploy LLMs on Foundry

Question

First I need to say I am very familiar with Model Catalog service on Foundry. I used it a lot.
But today, I noticed that some changes happens: I can not deploy a general LLM like gpt-5.1 on Foundry.
The error is the model "isn't available due to insufficient quota". I never see this message previously for the general models like gpt-5.1, etc.

can you explain the reason?

Answer 1

Hello @JingJing Bao ,

Welcome to Microsoft Q&A .Thank you for reaching out to us.

Thank you for sharing the details and screenshot. After reviewing the deployment behavior and the message shown in the deployment dialog, the issue appears to be occurring during quota/capacity validation for the selected deployment configuration, rather than during model discovery.

The model remains available in the Foundry Model Catalog; however, the selected deployment path is currently unable to obtain the required quota allocation or capacity needed to create the deployment.

One important observation is that the issue description references gpt-5.1, while the screenshot shows gpt-5.4-nano. Since quota is evaluated independently for each:

• Model and model version
• Deployment type
• Azure region
• Subscription quota allocation

confirming the exact model experiencing the issue will help determine whether the behavior is isolated to a specific model or affects multiple GPT models.

The screenshot provides an important diagnostic indication:

Default settings

• "Set to global standard and default quota"
• Appears unavailable

Custom settings

• "Set your own SKU, quota, PTU, spillover, and guardrails"
• Remains available

This indicates that the default deployment configuration is currently unable to satisfy the required quota/capacity requirements for the selected model and environment.

This does not necessarily mean that the model is unavailable or that all deployment options are blocked. Instead, it suggests that the limitation is related to the selected deployment configuration and available quota/capacity.

The availability of Custom settings indicates that alternative deployment configurations can be reviewed. Deployment success will still depend on whether sufficient quota and capacity are available for the selected configuration.The screenshot provides an important diagnostic indication:

Default settings

• "Set to global standard and default quota"
• Appears unavailable

Custom settings

• "Set your own SKU, quota, PTU, spillover, and guardrails"
• Remains available

This indicates that the default deployment configuration is currently unable to satisfy the required quota/capacity requirements for the selected model and environment.

This does not necessarily mean that the model is unavailable or that all deployment options are blocked. Instead, it suggests that the limitation is related to the selected deployment configuration and available quota/capacity.

The availability of Custom settings indicates that alternative deployment configurations can be reviewed. Deployment success will still depend on whether sufficient quota and capacity are available for the selected configuration.

For the previous deployments, the behavior does not necessarily indicate a change in Model Catalog availability.

This can occur because deployment validation evaluates the current quota and capacity conditions at the time of deployment. Availability may vary depending on:

• Selected model/version
• Deployment type
• Azure region
• Subscription quota allocation
• Existing deployments consuming quota
• Current available capacity

For example, quota allocated for an existing model deployment does not automatically provide quota for another model or model version.

A deployment that succeeded previously may fail later if:

• A different model/version is selected.
• Existing deployments consume the available quota allocation.
• The selected region has limited capacity for that model.
• The selected deployment type requires quota that is currently unavailable.

Foundry quota availability should therefore be evaluated based on the specific combination of model + region + deployment type + subscription

Please check if the following steps help-

1. Confirming the Deployment Details Please confirm the following:
   • Exact model name and version:
   • The issue description mentions GPT-5.1.
   • The screenshot shows GPT-5.4-nano.
   • Azure region where deployment is being attempted.
   • Deployment type selected:
   • Global Standard
   • Standard
   • Provisioned (if applicable)
   • Subscription type:
   • Pay-As-You-Go
   • Enterprise Agreement
   • CSP
   • Trial
   • Other subscription offers
   This information will help determine whether the issue is model-specific, region-specific, or subscription-related.
2. Testing Deployment Using Custom Settings Since Custom settings remains available in the deployment dialog:
   1. Open the deployment workflow.
   2. Select Custom settings instead of Default settings.
   3. Review the available deployment configurations.
   4. Check whether another supported deployment type is available.
   5. If the deployment experience allows, test with an appropriate TPM allocation.
   6. Verify whether the deployment completes successfully.
   This step helps determine whether the issue is limited to the default Global Standard deployment path or affects all available deployment options.
3. . Reviewing Model-Specific Quota Allocation Navigate to: Microsoft Foundry portal → Management Center → Quota Review the following:
   • Total quota
   • Allocated quota
   • Available quota
   • Existing deployments consuming quota
   • Affiliated deployments using the same quota pool
   If available quota is fully allocated:
   • Review existing deployments consuming quota.
   • Reduce TPM allocations for unused or oversized deployments where appropriate.
   • Remove unused deployments if they are no longer required.
   • Retry deployment after quota changes are reflected.
4. Validating Regional Availability Quota and capacity availability can vary by region. Recommended validation:
   • Attempt deployment in another supported region.
   • Compare the result using the same model and deployment type.
   If deployment succeeds in another region, this would indicate a regional capacity limitation rather than a model availability issue.
   1. Verifying access Permissions
   If quota information appears incomplete or unavailable, verify that appropriate permissions are assigned: Cognitive Services Usages Reader
   • Required to view quota information.
   Cognitive Services Contributor
   • Required for resource management operations.
   Owner or Contributor
   • Required for certain subscription-level quota management actions.
   Insufficient permissions may prevent accurate quota visibility within the portal.

The following references might be helpful , please check them out

• Deploy Microsoft Foundry Models in the Foundry portal - Microsoft Foundry | Microsoft Learn
• Manage and increase quotas for resources - Microsoft Foundry | Microsoft Learn
• Azure OpenAI in Microsoft Foundry Models Quotas and Limits - Microsoft Foundry | Microsoft Learn
• Understanding deployment types in Microsoft Foundry Models - Microsoft Foundry | Microsoft Learn
• Foundry Models sold by Azure - Microsoft Foundry | Microsoft Learn
• Role-based access control for Azure OpenAI (classic) - Microsoft Foundry (classic) portal | Microsoft Learn
• Microsoft Foundry Service: Request for Quota Increase

Please let us know if the response was helpful

 

Thank you

 

Answer 2

That message indicates quota is now being enforced for that model/region/deployment path, not that the model catalog is generally broken.

Supported reasons from the documentation:

1. Quota can block deployment Foundry deployment failures can occur with Out of quota or quota exceeded conditions. For managed compute deployments, deployment can fail when there is insufficient quota for the required VM SKU. The documented actions are to check current quota usage, request a quota increase for the needed capacity, or choose a different available SKU.
2. Quota is model- and region-dependent If quota for a specific model cannot be found or requested, regional availability must be checked because not all models are available in all regions.
3. Deployment type matters In Foundry Models, serverless API deployments and managed compute are different:
   • serverless API deployments are deployed within an AI project resource
   • managed compute deployments run on Azure Machine Learning managed compute and require subscription VM quota for the specific product/SKU
4. Some models require specific access/availability For Azure OpenAI models, a deployment can also fail if the subscription or region does not have access to that model.

Practical interpretation for the scenario shown:

• If the deployment path is managed compute, the most likely cause is missing quota for the required VM capacity.
• If the deployment path is a Foundry Models / Azure OpenAI-style deployment, the quota or availability can still vary by model and region, so a model that was previously easy to deploy may now show insufficient quota in the selected configuration.
• If the quota UI looks wrong or empty, the documented checks are:
  • confirm the correct subscription is selected
  • verify Cognitive Services Usages Reader at subscription scope to view quota
  • verify Owner or Contributor at subscription scope to request quota
  • allow up to 15 minutes for approved quota changes to propagate

If the goal is to confirm the reason quickly, the supported path is:

1. Check which deployment option is selected: Serverless API deployment or Managed compute.
2. In the Azure portal or Foundry quota experience, check quota for the selected model/region or required VM SKU.
3. If quota is unavailable, request more quota where supported, or try a different region/model/SKU.
4. If the quota page is empty, verify the required subscription-level roles.

---

References:

• How to troubleshoot your deployments and monitors in Microsoft Foundry portal (classic)
• How to deploy and infer with a managed compute deployment (classic) (python-sdk)
• Manage and increase quotas for resources with Microsoft Foundry (Foundry projects)
• Microsoft Foundry Models overview (classic)
• Microsoft Foundry Models overview