Anthropic model Claude Opus 4.6 deployment fails every time in Microsoft Foundry

Question

Anthropic model Claude Opus 4.6 deployment fails every time in Microsoft Foundry

Panagiotis Tzedakis 47

Hi everyone,

I am working on a project where we have to create a Foundry account, Foundry project and deploy Anthropic model Claude Opus 4.6 (GlobalStandard, 40K TPM).

I have the Contributor role in a new PAYG (Pay-as-you-go) Azure subscription and Marketplace access is enabled for the subscription. I navigate to Microsoft Foundry web site (https://ai.azure.com/<tenant>) and create a project in Azure region Sweden Central. Then, I clicke on Discover, find the Anthropic model Claude Opus 4.6 and I start the model deployment using the default settings. After a few minutes the model deployment reports status Failed without giving me any more context with regards to the exact reason it failed.

I have tried the exact same thing on another subscription and it works without any issues.

I am lost. Can anyone please help me locate the issue?

SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-05-21T15:40:26.5+00:00

Hi @Panagiotis Tzedakis

We are currently investigating the issue with the internal team. Please allow us some additional time, and we will share an update with you as soon as we have more information.

Thank you!
Panagiotis Tzedakis 47 Reputation points

2026-05-23T11:02:36.86+00:00

Hi, did you find anything with regards to my issue? Is there anything I can do to help?
Haig Didizian 1 Reputation point

2026-05-28T20:24:39.03+00:00
Hello,

I'm having the same issue across multiple subscriptions. I think the details are very similar to @Panagiotis Tzedakis :

Anthropic model (in this case, claude-haiku-4-5)

I have quota available for the model

Deployment says "creating" for 5+ minutes, then fails with just "Failed"

This is the same whether I deploy via CLI or Foundry Portal

For what it's worth, this appears to be happening with multiple Claude models.

I'm definitely interested in what you're able to find out - I'm stuck without any detailed error logging.

Thank you!

Haig Didizian

Clearistic Technology
SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-05-29T17:33:59.29+00:00

Hi @Panagiotis Tzedakis

Did you get any chance to review the above response. Do let me know if you have any further queries.

Thank you!
Panagiotis Tzedakis 47 Reputation points

2026-06-03T16:10:15.4466667+00:00

Hi @SRILAKSHMI C , I deleted all the model deployments and all foundry projects but one. I tried to create a new Claude Opus 4.6 model deployment today which succeeded.
SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-06-09T06:13:39.88+00:00

Hi @Panagiotis Tzedakis

Glad to hear it’s working now! Thanks for confirming the solution.

Since I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted?

Thank you!

Answer accepted by question author

kagiyama yutaka 5,570

I think region, Contributor on the Foundry resource group, the Anthropic Marketplace offer being allowed, and that the subscription can allocate the Anthropic GlobalStandard quota. If the quota is zero, the deployment just fails with no extra detail.

Panagiotis Tzedakis 47 Reputation points

2026-05-24T07:31:13.99+00:00

Hi Yuma and thanks for your comment.

At the time of creating the Microsoft Foundry resource there was available the whole 40K TPM quota to consume, so I do not think this is why the Microsoft Foundry deployment failed.
kagiyama yutaka 5,570 Reputation points

2026-05-24T08:24:31.1866667+00:00

I see, thanks. Just to clarify — 40K TPM shown in Foundry is the project‑side limit. The Global quota is configured in the Azure subscription per region, and deployments can fail if the required subscription quota is not available. Maybe worth checking the subscription quota.
Panagiotis Tzedakis 47 Reputation points

2026-05-24T09:02:05.6533333+00:00

I see. I did not know that. How can I check the subscription quota?
Panagiotis Tzedakis 47 Reputation points

2026-05-24T09:19:23.6366667+00:00

I checked the quota and the thing is that before I deployed the model in the specific region, I had all the quota available (40K TPM). When I attempted to deploy the model, the deployment failed but the subscription quota for the model now shows 0 TPM available, which I feel makes sense. However, I still do not understand why the deployment gets failed. It does not even log a message to help me understand the failure reason. It is strange.
kagiyama yutaka 5,570 Reputation points

2026-05-24T09:25:21.3033333+00:00

Global quota only inside Foundry → Management Center → Quota. If it shows 0/0, the subscription has no assigned capacity, and Azure support handles the “Anthropic Global Standard enablement” request.　
Panagiotis Tzedakis 47 Reputation points

2026-05-24T19:01:13.8033333+00:00

Thank you. @SRILAKSHMI C Can you check what Yutaka says in case this is something you need to handle on your end? My client needs to have the model setup asap and I do not know what else to do.
kagiyama yutaka 5,570 Reputation points

2026-05-27T02:57:30.2833333+00:00

Thanks. Yes — this is a subscription‑side GlobalStandard enablement. SRILAKSHMI can assign the quota on their side.

1 additional answer

Your answer

SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-05-21T15:40:26.5+00:00

Hi @Panagiotis Tzedakis

We are currently investigating the issue with the internal team. Please allow us some additional time, and we will share an update with you as soon as we have more information.

Thank you!
Panagiotis Tzedakis 47 Reputation points

2026-05-23T11:02:36.86+00:00

Hi, did you find anything with regards to my issue? Is there anything I can do to help?
Haig Didizian 1 Reputation point

2026-05-28T20:24:39.03+00:00

Hello,

I'm having the same issue across multiple subscriptions. I think the details are very similar to @Panagiotis Tzedakis :

Anthropic model (in this case, claude-haiku-4-5)

I have quota available for the model

Deployment says "creating" for 5+ minutes, then fails with just "Failed"

This is the same whether I deploy via CLI or Foundry Portal

For what it's worth, this appears to be happening with multiple Claude models.

I'm definitely interested in what you're able to find out - I'm stuck without any detailed error logging.

Thank you!

Haig Didizian

Clearistic Technology
SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-05-29T17:33:59.29+00:00

Hi @Panagiotis Tzedakis

Did you get any chance to review the above response. Do let me know if you have any further queries.

Thank you!
Panagiotis Tzedakis 47 Reputation points

2026-06-03T16:10:15.4466667+00:00

Hi @SRILAKSHMI C , I deleted all the model deployments and all foundry projects but one. I tried to create a new Claude Opus 4.6 model deployment today which succeeded.
SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-06-09T06:13:39.88+00:00

Hi @Panagiotis Tzedakis

Glad to hear it’s working now! Thanks for confirming the solution.

Since I’ve converted my earlier comment into an answer, could you please take a moment to mark it as Accepted?

Thank you!
Panagiotis Tzedakis 47 Reputation points

2026-05-24T07:31:13.99+00:00

Hi Yuma and thanks for your comment.

At the time of creating the Microsoft Foundry resource there was available the whole 40K TPM quota to consume, so I do not think this is why the Microsoft Foundry deployment failed.
kagiyama yutaka 5,570 Reputation points

2026-05-24T08:24:31.1866667+00:00

I see, thanks. Just to clarify — 40K TPM shown in Foundry is the project‑side limit. The Global quota is configured in the Azure subscription per region, and deployments can fail if the required subscription quota is not available. Maybe worth checking the subscription quota.
Panagiotis Tzedakis 47 Reputation points

2026-05-24T09:02:05.6533333+00:00

I see. I did not know that. How can I check the subscription quota?
Panagiotis Tzedakis 47 Reputation points

2026-05-24T09:19:23.6366667+00:00

I checked the quota and the thing is that before I deployed the model in the specific region, I had all the quota available (40K TPM). When I attempted to deploy the model, the deployment failed but the subscription quota for the model now shows 0 TPM available, which I feel makes sense. However, I still do not understand why the deployment gets failed. It does not even log a message to help me understand the failure reason. It is strange.
kagiyama yutaka 5,570 Reputation points

2026-05-24T09:25:21.3033333+00:00

Global quota only inside Foundry → Management Center → Quota. If it shows 0/0, the subscription has no assigned capacity, and Azure support handles the “Anthropic Global Standard enablement” request.　
Panagiotis Tzedakis 47 Reputation points

2026-05-24T19:01:13.8033333+00:00

Thank you. @SRILAKSHMI C Can you check what Yutaka says in case this is something you need to handle on your end? My client needs to have the model setup asap and I do not know what else to do.
kagiyama yutaka 5,570 Reputation points

2026-05-27T02:57:30.2833333+00:00

Thanks. Yes — this is a subscription‑side GlobalStandard enablement. SRILAKSHMI can assign the quota on their side.

Answer 1

Hi @Panagiotis Tzedakis

Thank you for the additional details.

What Yutaka mentioned is correct for Anthropic Global Standard deployments, there are two separate layers involved:

The quota visible within Azure AI Foundry

The backend subscription/regional quota allocation required by the Azure OpenAI service

In some cases, the Foundry UI may display available TPM quota, but the backend provider enablement or regional capacity allocation for the subscription may not yet be fully provisioned. This can result in deployments failing without a detailed validation message.

Since:

The same deployment succeeds in another subscription

Marketplace access is already enabled

You had visible quota prior to deployment

The quota changes after deployment attempts

this may indicate a backend quota allocation or provider onboarding issue for the subscription.

At this stage, we are checking this internally with the engineering team to verify:

Subscription-level Anthropic enablement
Regional quota allocation for Sweden Central
Backend deployment validation logs

As a temporary workaround, you may also test:

A smaller TPM deployment

Another supported region

Recreating the deployment after a few minutes in case quota propagation is still in progress

We’ll investigating this from our side and share updates as soon as we have more information.

Thank you!