Will Azure Foundry offer deployable nano GPT 5.x models?

Question

https://learn.microsoft.com/en-us/azure/foundry/foundry-models/concepts/models-sold-directly-by-azure-region-availability?tabs=az-americas&pivots=batch says gpt-4.1-nano batch is available, but I can't deploy it because it says that model is deprecated.

ServiceModelDeprecating

The model 'Format:OpenAI,Name:gpt-4.1-nano,Version:2025-04-14' is in deprecating state and cannot be used for new deployments.

Will there be a deployable nano batch replacement?

Answer 1

Hi Jason,

Yes, lightweight nano models in the GPT series are designed primarily for Global Batch Deployments rather than real-time provisioned endpoints due to regional GPU allocation constraints.

If a nano model shows as available in the Azure AI Foundry catalog documentation but fails during endpoint creation with a "region unavailable" error, it means the model is enabled for Batch Processing Jobs in that region rather than real-time latency deployment.

Recommended Workarounds:

1. Switch to Batch Inference: Use the Azure AI Foundry Batch API (/v1/chat/completions with batch parameters) instead of real-time endpoint deployment.
2. Deploy in Alternate Regions: Test provisioning in Sweden Central or West US 3 where real-time provisioned throughput for lightweight models is actively supported.
3. Use GPT-4o-mini for Real-Time: If you need low-cost, low-latency real-time inference today, deploy gpt-4o-mini which has full regional real-time deployment support.

Ref: Azure AI Foundry Model Catalog & Regional Availability

If this helps clarify model deployment availability, please mark as Accept Answer!

Answer 2

Hello @Jason X ,

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

The behavior observed appears to be related to the model lifecycle state rather than regional Batch availability.

The availability matrix and lifecycle policy are designed to provide different types of information , so reviewing them together explains why a model can remain visible in a Batch availability table while deployment creation is restricted.

Regarding the availability page show GPT-4.1-Nano Batch as available while deployment creation fails - The Region Availability matrix indicates where a model is supported based on deployment type and region, while the Model Lifecycle policy governs whether new deployments can be created for that model version.

The error: ServiceModelDeprecating"The model 'Format:OpenAI,Name:gpt-4.1-nano,Version:2025-04-14' is in deprecating state and cannot be used for new deployments."indicates that gpt-4.1-nano (Version: 2025-04-14) has entered the Deprecated lifecycle stage.

During this stage, eligibility for creating deployments is evaluated at the subscription level and a different subscription within the same tenant does not automatically inherit access. As a result, the model can still appear in the Batch availability matrix while deployment creation is restricted by lifecycle policy.

Regarding if GPT-4.1-Nano deprecated - Yes. The current retirement schedule lists:

• Model: gpt-4.1-nano
• Version: 2025-04-14
• Lifecycle state: Deprecated
• Retirement date: October 14, 2026
• Replacement: Not currently specified

At present, no official replacement model has been announced in the retirement schedule

Regarding will deployable GPT-5.x Nano models be available - GPT-5 Nano models are already available today through supported Standard deployment types. Current availability includes:

• gpt-5-nano (2025-08-07) through Global Standard and Data Zone Standard deployments.
• gpt-5.4-nano (2026-03-17) through Global Standard and Data Zone Standard deployments.

However, based on the currently published Batch deployment matrices:

• gpt-5-nano is not listed for Global Batch deployments.
• gpt-5.4-nano is not listed for Global Batch deployments.
• Neither model is currently listed for Data Zone Batch deployments.

At this time, there is no publicly documented release timeline or published commitment for GPT-5 Nano Batch support.

Regarding if there's a GPT-5.x Nano replacement for Batch deployments

Based on the current documentation:

• No GPT-5 Nano model is currently available for Global Batch or Data Zone Batch deployments.
• No official GPT-5 Nano Batch replacement has been announced for gpt-4.1-nano.

If Batch processing remains a requirement, gpt-5.4-mini is one GPT-5-family Batch-capable model that may be evaluated based on workload requirements. This should be considered a potential migration option and not an officially designated replacement.

As of now , the current retirement schedule does not identify an official replacement for gpt-4.1-nano, and the Replacement field remains unspecified. Future Batch model availability should continue to be validated against the latest Azure AI Foundry documentation.

The following references might be helpful , please check them out

• Region availability for Foundry Models sold by Azure - Microsoft Foundry | Microsoft Learn
• Foundry Models lifecycle and support policy - Microsoft Foundry | Microsoft Learn
• Model retirement schedule - Microsoft Foundry | Microsoft Learn
• How to use global batch processing with Azure OpenAI in Microsoft Foundry Models - Microsoft Foundry | Microsoft Learn

Please let us know if the response was helpful

 

Thank you