Azure OpenAI gpt40 auto-upgrade and retirement

Question

We are trying to determine if the Azure Open AI gpt4o (version: 2025-08-06) retirement and auto-upgrade dates.

Based on this (Azure OpenAI in Microsoft Foundry Model Retirements - Azure OpenAI | Microsoft Learn) document , it appears that it is going to be retired on 2026-10-01, but I wanted to confirm if the ‘Data Zone Standard’ deployment of gtp4o model will also to auto-upgrade in March or it won’t be (same as ‘Standard’ deployment)?

And if we upgrade to the gpt4o model version ‘2025-11-20’, then we won’t be impacted until Oct 2026 in relates t retirement as well as auto-upgrade?

 

Reference: Azure OpenAI in Microsoft Foundry Model Retirements - Azure OpenAI | Microsoft Learn

 

Answer 1

Hi Rahul Shah,

This behavior is expected and is tied to how Azure OpenAI model versioning and upgrade policies work.

In Azure OpenAI, deployments are created against a model family (for example, GPT‑4) and can be configured to either:

• Automatically move to the latest default version, or
• Stay pinned to a specific version until that version is retired.

If your GPT‑4.0 deployment was created with the default (auto‑update) policy, Azure will automatically upgrade it when:

• A newer version becomes the default, or
• The currently deployed version reaches its retirement date.

Microsoft documents this behavior clearly: Azure notifies customers ahead of time when a new version becomes the default, and older versions are kept available only until their scheduled retirement. Once retired, deployments are automatically moved forward to a supported version to avoid service disruption. [techcommun...rosoft.com]

What you can do going forward

• If you need strict behavior consistency, deploy a specific model version and choose an update policy that does not auto‑upgrade.
• Plan for periodic validation, as response quality and behavior can change slightly between model versions.
• Keep an eye on the official model retirement and upgrade announcements so you can test newer versions before they become mandatory.

In short: Nothing is “wrong” with your deployment this is standard lifecycle management for Azure OpenAI models. Pin versions when needed and treat upgrades as part of normal operational planning.

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

Thankyou!