GPT 4o and GPT 4.1 were deprecated today

Hello @Eugenio Segala ,

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

At this stage, the behavior appears related to model lifecycle enforcement and deployment-type availability constraints

The deployment requests are currently failing with ServiceModelDeprecating: The model is in deprecating state and cannot be used for new deployments.

This error indicates that the service is treating the selected model/version as part of the lifecycle deprecation process. According to the published lifecycle policy, Deprecated and Retired are different lifecycle stages. During the deprecated stage, restrictions can be introduced for deployment creation, while retirement occurs at a later date when the model reaches end of service. Existing deployments are generally expected to continue operating until retirement.

We understand the main concerns raised that these models were expected to remain available until October. The currently published retirement schedule still shows future retirement dates for the affected models:

• GPT-4o (2024-11-20) – retirement date: 2026-10-01
• GPT-4.1, GPT-4.1-mini, and GPT-4.1-nano (2025-04-14) – retirement date: 2026-10-14

Because deployment restrictions are being observed before those dates, the behavior appears likely to be related to lifecycle enforcement and deployment availability controls.

Another important consideration is subscription-level eligibility. The lifecycle policy states that deprecated models remain available to existing customers, while new customers cannot create deployments. Existing-customer status is determined at the subscription level and depends on whether that subscription has previously deployed the specific model version. Since these models were reportedly deployed previously, confirming whether the affected subscription previously deployed the same model/version is an important validation step.

We understand the concern regarding replacement options is also understood. The challenge is not only identifying a newer model, but identifying one that:

1. Is available in UK South
2. Supports the required deployment type
3. Meets data residency requirements
4. Maintains a comparable cost profile

Model availability varies by deployment type, including Standard (Regional), Global Standard, Data Zone Standard, and Provisioned Throughput (PTU). As a result, a model may be available in one deployment type but not in another.

Please check if the following steps help to determine whether this is lifecycle enforcement, deployment-type availability, or a subscription-specific eligibility issue

1. Please confirm the exact:
   • Model name
   • Model version
   • Region
   • Deployment type
2. Check and confirm whether the same subscription previously deployed the exact model/version being redeployed.
3. Please verify whether the issue affects:
   • New deployments only or
   • Existing active deployments as well.
4. Test deployment through REST API or CLI in addition to Azure AI Foundry and compare the results.
5. Please validate model availability across:
   1. Standard (Regional)
   2. Global Standard
   3. Data Zone Standard
   4. Provisioned Throughput (PTU)
6. Consider comparing the model lifecycle state returned through the Models API with the published retirement schedule, including lifecycle status and deprecation information

As a precaution, deleting any currently functioning deployments is not recommended until the lifecycle status, subscription eligibility and replacement availability have been fully validated.

The following references might be helpful , please check them out

• Foundry Models lifecycle and support policy - Microsoft Foundry | Microsoft Learn
• Understanding deployment types in Microsoft Foundry Models - Microsoft Foundry | Microsoft Learn
• Region availability for Foundry Models sold by Azure - Microsoft Foundry | Microsoft Learn
• Model retirement schedule - Microsoft Foundry | Microsoft Learn

Please let us know if the response was helpful

Thank you

Thank you @Karnam Venkata Rajeswari ,

All these models gpt-4o, gpt-4o-mini, gpt-4.1, gpt-4.1-mini and gpt-4.1-nano are basically gone from Azure Open AI.

The error is the same across all of them:

"Failed to deploy model gpt-4.1-mini ServiceModelDeprecating: The model 'Format:OpenAI,Name:gpt-4.1-mini,Version:2025-04-14' is in deprecating state and cannot be used for new deployments."

To be clear, no alternatives are present because GPT 5 and other models are not available in UK South only using the "Standard" deployment model. GPT 5.4 can be deployed via PTUs, but costs 20x what gpt-4.1-mini costed us, using the same deployment model.

gpt-5.4-mini is not also available in the UK only deployments.

Leaving us with no alternatives really.

It's worth pointing out that the original retirement date for those models was October.

Hello @Karnam Venkata Rajeswari ,

Did you have a chance to look into the above?

I believe it's too early to remove those models from Azure, as there aren't equivalent alternatives for UK users with data restrictions.

Our current use-case is at risk.

To add more context, we have deployed those models several times across our subscriptions.

But now we are unable to re-deploy them under the same ones.

@Karnam Venkata Rajeswari what you are suggesting should not be applicable to us, as we deployed those models before in our subscriptions.

This seems to be a bug.