Can We extend the current Azure OpenAI Model Deployment in MSFT Foundry/Azure OpenAI

Question

Can We extend the current Azure OpenAI Model Deployment in MSFT Foundry/Azure OpenAI

Lu, Isabel (she/her/hers) 20

Our organization has deployed the GPT‑4.1 model and manages it through Microsoft Foundry. We understand that the official retirement date for this model is October 14, 2026.

We want to learn if it is possible to request an extension to the retirement date for this model? If so, could you please advise on the process for requesting such an extension? Thanks.

SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-04-10T17:08:00.6233333+00:00
Hello Lu, Isabel (she/her/hers),

Thank you for reaching out regarding the GPT-4.1 model lifecycle in Azure OpenAI / Microsoft Foundry.

At a high level, retirement dates for Azure OpenAI models are fixed at the platform level, and cannot be extended on a per-customer basis. For GPT-4.1, the October 14, 2026 retirement date is firm, and workloads will need to be migrated before that point.

Can the retirement date be extended?

In general Model retirement timelines are not extendable per customer, as they are managed centrally for platform reliability, security, and consistency.

Key dates for GPT-4.1

Deprecation date: April 14, 2026

Retirement date: October 14, 2026

After the retirement date:

The model will no longer serve inference requests

Existing deployments will stop working

You must transition to a supported model

Recommended approach moving forward

Since extensions are not typically available, the focus should be on controlled migration and minimizing disruption.

Configure an update policy

In Microsoft Foundry deployments, you can define how your deployment behaves when a model is retired:

Auto-update to default → Automatically moves to the latest default model version designated by Azure

Upgrade when expired → Keeps using GPT-4.1 until retirement, then switches to a successor

This ensures:

No sudden downtime

Automatic fallback to a supported model (for example, GPT-4o or newer equivalents)

Plan and validate your migration path

You should begin testing replacement models early.

Review available models and regions: https://learn.microsoft.com/azure/ai-foundry/openai/concepts/models#model-summary-table-and-region-availability

Validate in dev/test:

Output quality

Latency

Token usage

Feature compatibility

Run parallel deployments

To reduce risk:

Keep GPT-4.1 running

Deploy a successor model alongside it

Gradually shift traffic

This allows Controlled validation, Easy rollback if needed

Avoid hardcoding model names

Instead of hardcoding:

Use configuration-based model selection so you can:

Switch models without code changes

Adapt quickly to future updates

Be aware of automatic transitions

Depending on your deployment configuration:

Some deployments may auto-upgrade near retirement

Behavior depends on your update policy and deployment type

It’s recommended to test this behavior in advance to avoid surprises.

Please refer this

Model deprecations & retirements: https://learn.microsoft.com/azure/ai-services/openai/concepts/model-retirements

Working with model update policies: https://learn.microsoft.com/azure/ai-services/openai/how-to/working-with-models?tabs=powershell#model-updates

Model availability by region: https://learn.microsoft.com/azure/ai-foundry/openai/concepts/models#model-summary-table-and-region-availability

I Hope this helps. Do let me know if you have any further queries.

Thank you!
SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-04-14T17:32:01.53+00:00

Hi Lu, Isabel (she/her/hers),

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

Thank you!

1 answer

Your answer

SRILAKSHMI C 19,735 Reputation points Microsoft External Staff Moderator

2026-04-14T17:32:01.53+00:00

Hi Lu, Isabel (she/her/hers),

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

Thank you!

Answer 1

An extension to the official retirement date of an Azure OpenAI model (including GPT‑4.1 managed via Microsoft Foundry/Azure AI Foundry) cannot be requested on a per‑customer basis.

Model retirement and upgrade timelines are centrally defined and enforced by the service:

Model availability and lifecycle (Preview → GA → Legacy → Deprecated → Retired) are managed globally by Microsoft and model providers, not by individual customer requests.
When a model reaches the Retired stage, existing deployments stop working and calls return errors; there is no supported mechanism to keep a specific deployment running beyond the published retirement date.
For Azure OpenAI in Foundry/Azure AI Foundry, retirement is done on a rolling, region‑by‑region basis, but the dates and behavior are still centrally controlled and not extendable per tenant or subscription.
Customers are expected to migrate to the designated replacement model before the retirement date. For GPT‑4.1 and other models, the retirement and replacement mapping is documented in the Azure OpenAI model deprecations and retirements pages.

What can be done instead:

Plan and execute migration before retirement
Follow the guidance under “Preparation for model retirements and version upgrades” to:
- Test applications against the recommended replacement model.
- Update deployments and application configuration to use the new model/version before the retirement date.
Monitor retirement notifications
Ensure that:
- Subscription owners and anyone with Reader or above permissions monitor Azure Resource Health and configure Service Health alerts for Azure OpenAI service with event types Upgrade, Deprecation, & Retirement Notifications.
- Email notifications sent to subscription owners are being received and acted upon.
Use support only for clarification, not extensions
A support request can clarify timelines or migration behavior, but it cannot be used to obtain an extension or a custom exception to the retirement date.

In summary, the retirement date for GPT‑4.1 in Foundry/Azure OpenAI is fixed at the service level and cannot be extended for a specific organization. The supported path is to migrate to the recommended successor model before the published retirement date.

References: