Incomplete 'ModelProviderData' input fields in Azure AI Foundry UI for Anthropic Claude Deployment

Question

claude fail.pdf

Unfortunately, the same issue persists even after following the exact steps you provided below.

• Issue: Claude model deployment fails in the Microsoft AI Foundry portal --> As a result, the Claude model provisioning cannot be completed.

The steps previously provided:

1. Open Azure Portal --> Search for "AI Foundry"

Create/select a Foundry project (supported regions: East US2 or Sweden Central)

1. Go to Model catalog --> Search for "Claude"
2. Select your model (e.g., claude-sonnet-4-8) --> Click Deploy
3. Choose Global Standard deployment --> Confirm

I have attached the screenshots capturing each step in chronological order:

step 1.jpg: Searching and accessing Microsoft Foundry in the Azure Portal.

step 2.jpg: Verifying the resource (dev-1141-resource) and region (eastus2).

step 3.png: Viewing the Deployments page in the Foundry portal where other models have successfully deployed.

step 4.jpg: Configuring the deployment settings for claude-opus-4-8 using the 'Global Standard' deployment type.

step 5.png: The final Deployments list showing that while other models (like GPT-5.5) succeeded, the claude-opus-4-8 deployment status explicitly shows "Failed".

Since this needs to be integrated into our company's shared internal automation pipeline, we cannot utilize localized or alternative standalone workarounds. Successful model provisioning is our highest priority right now, but we are completely blocked at the initial provisioning stage.

Please review this deployment failure and advise on a solution as soon as possible.

Answer 1

Hello @volt-ai ,

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

Deployment attempts for Anthropic Claude models are consistently reaching a Failed state, while other models deploy successfully under the same setup.

This confirms that:

• The Foundry resource, region, and base configuration are valid
• The issue is specific to the Anthropic Claude deployment path

Based on analysis and validated scenarios, this behavior is most often caused by one or more of the following:

1. ncomplete or missing modelProviderData fields - (industry, organizationName, countryCode not fully passed during deployment)
2. Backend provisioning failure during:
3. Marketplace subscription creation
   1. Anthropic organization binding
4. Tenant metadata gaps, where required fields are not returned by the platform and fail validation

This means the issue is not due to incorrect deployment steps, but rather due to provider metadata handling or backend provisioning inconsistencies.

Please check if the following validation and troubleshooting steps help-

1. Region validation Please confirm project is created in:
   • East US 2
   • Sweden Central
2. Subscription & Access Validation
   1. Ensure a supported paid subscription is used
   2. Confirm Marketplace access is enabled
3. Deployment Retry Best Practices
   1. Create a new deployment name (avoid reusing failed name)
   2. Optionally delete failed deployment and recreate
   3. Retry after 15–30 minutes to allow backend synchronization
   4. Verify required roles (Contributor / Owner)
4. Project Isolation Test
   1. Create a new Foundry project
   2. Attempt deployment in fresh environment

To bypass UI-level inconsistencies, deployment via REST/CLI is recommended, as it allows explicit definition of required fields.

1. Preparing payload

      {
        "sku": {
          "name": "GlobalStandard",
          "capacity": 1
        },
        "properties": {
          "model": {
            "format": "Anthropic",
            "name": "claude-opus-4-8",
            "version": "1"
          },
          "modelProviderData": {
            "industry": "technology",
            "organizationName": "Contoso Ltd",
            "countryCode": "US"
          }
        }
      }
   

2. Deploy using CLI

      az rest --method put \
        --url "https://management.azure.com/subscriptions/<subscription-id>/resourceGroups/<resource-group>/providers/Microsoft.CognitiveServices/accounts/<account-name>/deployments/<deployment-name>?api-version=2024-10-01" \
        --headers "Content-Type=application/json" \
        --body @deployment.json
   

Please note to

1. Use lowercase values for industry
2. Ensure all three fields are explicitly present

If the same failure persists after:

• clean deployment attempt
• REST/CLI deployment
• validation of subscription and region

then the issue should be treated as a backend provisioning problem, requiring investigation into:

• provider metadata population
• Marketplace integration workflow
• Anthropic organization creation step

We have reached out to you on private messenger to assist you further.

The following references might be helpful , please check them out

• Deploy and use Claude models in Microsoft Foundry - Microsoft Foundry | Microsoft Learn

Please let us know if the response was helpful

Thank you

Answer 2

The deployment failure for claude-opus-4-8 in Azure AI Foundry aligns with a general class of provisioning issues where the API deployment for the resource never reaches a Succeeded state.

From similar issues, there are two practical, supported actions:

1. Delete and recreate the Claude deployment
   • In the Foundry portal, go to the Deployments page for the dev-1141-resource project.
   • Delete the failed claude-opus-4-8 deployment.
   • Wait a short period, then redeploy the same model (claude-opus-4-8) again using Global Standard in a supported region (East US2 or Sweden Central).
   • This pattern has resolved stuck/failed provisioning states for Claude and other managed resources where the portal reports that the deployment is not ready.
2. If repeated failures persist, switch to a fresh deployment instance
   • If the same deployment name continues to fail after recreation, create a new deployment with a different name for the same model in the same Foundry project and region.
   • Ensure the subscription and region meet the Claude requirements (Enterprise or MCA‑E subscription and supported region). If other models (like GPT‑5.5) deploy successfully in the same project, this isolates the issue to the specific Claude deployment instance.

If the deployment repeatedly fails even after delete/recreate, the next step is to open a support ticket with Azure, referencing the failed deployment ID and time window, so the platform team can inspect the backend provisioning logs for that specific Claude deployment.

---

References:

• Deploy and use Claude models in Microsoft Foundry (preview)
• How to deploy and infer with a managed compute deployment (classic) (ai-foundry-portal)
• Deployed Claude and getting API deployment resource is not ready even though it has been an hour - Microsoft Q&A ERR_BAD_REQUEST
• Stuck in Creating / Failed provisioning Redis Cache - Microsoft Q&A
• Azure Deployment Takes 70 Minutes to Fail with Obscure Error - Microsoft Q&A