Claude Sonnet 4.6 Deployment Failure on Azure AI Foundry EastUS2

Question

Claude Sonnet 4.6 Deployment Failure on Azure AI Foundry EastUS2

Nutpapon Limsupavanich 0

Hi,

We attempted to deploy Claude Sonnet 4.6 in Azure AI Foundry, but the deployment failed with a backend service error. Deployment of Claude Sonnet 4.6 failed in Azure AI Foundry during deployment creation. The operation reached a terminal provisioning state of Failed and returned an internal server error from the Azure Cognitive Services deployment service. This appears to be a service-side failure (InternalServerError) rather than a configuration or authorization issue, based on the activity log. Could you please investigate the failure and advise on available alternatives?

Bests,

Nutpapon Lim

write-deployment.txt

Karnam Venkata Rajeswari 5,255 Reputation points Microsoft External Staff Moderator

2026-06-02T05:10:05.28+00:00
Hello @Nutpapon Limsupavanich ,

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

Thank you for sharing the detailed information. The observed behavior shows that the deployment successfully passed initial validation and then failed during the provisioning phase with an InternalServerError. This pattern generally indicates that the request reached backend allocation or partner onboarding steps, rather than failing due to authentication or permission issues.

The failure occurs during provisioning, which suggests either a backend service condition or an onboarding / metadata validation issue.

Please check if the following steps help -

Confirming subscription and deployment eligibility Claude models require a supported setup for successful provisioning.

Verifying that the resource is created in a supported region:

East US 2 (current region)

Sweden Central (alternate region)

Confirming that the subscription has:

A valid paid billing method

Access to partner model deployments via marketplace

Required permissions to create deployments

Validating quota and capacity availability Provisioning may fail if quota is unavailable or insufficient. Navigate to Foundry > Quotas Verify

Non-zero quota for Claude Sonnet models

No exhaustion of deployment slots or token limits

If quota is unavailable, a quota increase request may be required.

Reviewing provisioning and onboarding signals from Activity Logs The generic error may hide more specific failure messages.

Open Activity Log at the resource level

Filter using the operation or correlation ID and look for

AnthropicOrganizationCreationFailed

Marketplace or provisioning-related failures

These signals help determine whether the issue is backend-related or onboarding-related.

Validating provider metadata and onboarding requirements Claude deployments may require additional provider-specific details. Ensure the following fields if applicable are valid:

industry

organizationName

countryCode

Incorrect or missing values may result in provisioning failures presenting as InternalServerError.

. Retrying deployment to rule out transient conditions Provisioning failures can sometimes be temporary,so please consider

Deleting the failed deployment (if present)

Waiting briefly (10–30 minutes)

Retrying deployment

If the issue persists in the same region, attempt deployment in Sweden Central as a supported alternative.

The following references might be helpful , please check them out

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

Azure OpenAI in Microsoft Foundry Models Quotas and Limits - Microsoft Foundry | Microsoft Learn

Region availability for models in standard deployments - Azure Machine Learning | Microsoft Learn

Please let us know if the response was helpful

Thank you
Nutpapon Limsupavanich 0 Reputation points

2026-06-03T15:27:15.0233333+00:00

Hi Karnam Venkata Rajeswari

Thank you for your reply, I have reviewed and tested further in both East US 2 and Sweden Central.

All Claude model deployments fail during provisioning. This appears more likely to be related to subscription enablement or Anthropic backend account provisioning rather than a regional capacity issue.

For reference, I am using a personal Azure subscription with credit card billing (not CSP).

I also tested deploying GPT-5, GPT-4, and Omni models. I can't click deployment because they consistently show "Insufficient quota" in every region is list. In contrast, the Anthropic models allow me to start the deployment process, but the deployment ultimately fails during provisioning.

Is there a way to confirm whether my subscription has sufficient quota and access for Anthropic models? Specifically, is there any portal view, API, or backend validation that can confirm whether the deployment failures are caused by quota/access limitations versus Anthropic account provisioning issues?

Thank you.
Karnam Venkata Rajeswari 5,255 Reputation points Microsoft External Staff Moderator

2026-06-04T04:12:33.8433333+00:00

Hello @Nutpapon Limsupavanich ,

Thank you for assisting us with additional information. We will let you know the updates regarding this over the private messenger.

Thank you

1 answer

Your answer

Nutpapon Limsupavanich 0 Reputation points

2026-06-03T15:27:15.0233333+00:00

Hi Karnam Venkata Rajeswari

Thank you for your reply, I have reviewed and tested further in both East US 2 and Sweden Central.

All Claude model deployments fail during provisioning. This appears more likely to be related to subscription enablement or Anthropic backend account provisioning rather than a regional capacity issue.

For reference, I am using a personal Azure subscription with credit card billing (not CSP).

I also tested deploying GPT-5, GPT-4, and Omni models. I can't click deployment because they consistently show "Insufficient quota" in every region is list. In contrast, the Anthropic models allow me to start the deployment process, but the deployment ultimately fails during provisioning.

Is there a way to confirm whether my subscription has sufficient quota and access for Anthropic models? Specifically, is there any portal view, API, or backend validation that can confirm whether the deployment failures are caused by quota/access limitations versus Anthropic account provisioning issues?

Thank you.
Karnam Venkata Rajeswari 5,255 Reputation points Microsoft External Staff Moderator

2026-06-04T04:12:33.8433333+00:00

Hello @Nutpapon Limsupavanich ,

Thank you for assisting us with additional information. We will let you know the updates regarding this over the private messenger.

Thank you

Answer 1

The activity log description and InternalServerError indicate a backend/service-side issue with the Claude Sonnet 4.6 deployment rather than a configuration, authorization, or quota problem.

Based on the available guidance, use the following checks and alternatives:

Validate basic deployment prerequisites
- Confirm the deployment is in a supported region for Claude models. Claude models are only available in East US2 and Sweden Central.
- Ensure the subscription type is supported (Claude models require an Enterprise or MCA‑E subscription) and that you have Contributor or Owner on the resource group and Cognitive Services User for Entra ID access if using token-based auth.
Rule out common client-side causes
- Check for typical API errors (401, 403, 404, 429) as described in the Claude troubleshooting table. In this case, the activity log already shows InternalServerError at the Cognitive Services deployment layer, which points away from these client-side issues and toward a platform problem.
Check quota and capacity
- In the Foundry portal, verify quota under Operate > Quotas for Claude Sonnet in East US2. Ensure:
  - Quota is not exceeded for the model/region.
  - There is no regional capacity or quota error blocking new deployments.
- If quota is exceeded, either reallocate TPM from existing deployments or request more quota using the quota increase process referenced in the Foundry Models documentation.
Retry the deployment
- Because this is an internal service error, it may be transient. Delete the failed deployment (if it exists in a failed state) and attempt to create the deployment again after some time.
- If the deployment remains stuck in a non‑Succeeded provisioning state for an extended period, delete and recreate it, similar to the guidance from other deployment issues where recreating the deployment resolved the problem.
Try an alternative region or model as a workaround
- If repeated attempts in East US2 continue to fail with InternalServerError, try:
  - Deploying Claude Sonnet 4.6 in Sweden Central, the other supported region for Claude models.
  - Deploying another Claude model (for example, Haiku or Opus) in East US2 to see if the issue is specific to Sonnet in that region.
- If another model or region succeeds, this further confirms a regional/model‑specific backend issue.
When to escalate
- If:
  - Quota and permissions are correct,
  - The region is supported (East US2),
  - Multiple retries still fail with InternalServerError, and
  - Alternative regions/models are not viable for the workload,
- Then a backend investigation is required. Open a support ticket with Azure Support, including:
  - Subscription ID, resource name, region (East US2),
  - Deployment name and timestamp,
  - The full error details from the activity log and any deployment logs.
- This is similar to other quota/capacity and deployment issues where backend allocation or service state must be corrected by the platform team.

These steps allow confirming that configuration, quota, and region are valid, using another region/model as a workaround, and then escalating to Azure Support for a service-side fix if InternalServerError persists.

References: