Azure AI Foundry Sora 2 retirement date feels too early and conflicts with Azure communication — can support be extended to Sept 24?

Hello @Alex ,

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

The difference between the retirement timeline shown in Azure AI Foundry and earlier public communications is understandable and relates to how model lifecycles are managed at different levels of detail.

Azure AI Foundry manages models through a versioned lifecycle framework, where each model version progresses through defined stages:

• Preview > Generally Available (GA) > Deprecated > Retired

The retirement date shown in the portal applies to a specific deployed version, in this case:

• Sora 2 (version 2025-10-06)

This version-level timeline is the primary planning reference for deployments, because once a version reaches the Retired stage, all inference requests will stop working and return errors

The difference between the portal date (June 2, 2026) and earlier public messaging (Sept 24, 2026) is due to differences in scope and abstraction level:

• Public communications generally describe overall model availability windows
• Azure AI Foundry enforces version-specific lifecycle timelines
• Upstream or external timelines reflect broader model family support, not individual deployed versions

As a result, both timelines can coexist, but they apply to different layers of the service.

Therefore, from an operational standpoint please consider the portal retirement date as the current planning baseline for this specific deployment.

It is important to note that preview models can be retired without a direct replacement, which makes this a valid planning consideration.

To manage this situation effectively, the following steps are recommended:

1. Monitoring model updates - Please regularly review the Azure AI Foundry model catalog for new Sora versions or replacements
2. Validating lifecycle programmatically - Use the Models API to confirm lifecycle status, deprecation, and retirement dates.

      GET https://management.azure.com/subscriptions/{sub}/providers/Microsoft.CognitiveServices/locations/{location}/models?api-version=2024-10-01
   

3. Configuring proactive notifications - Enable Service Health alerts to receive updates on deprecations and retirements
4. While planning for continuity ,avoid reliance on a single version. Prepare fallback or alternative approaches if no replacement becomes available

The following references might be helpful , please check them out

• Foundry Models lifecycle and support policy - Microsoft Foundry | Microsoft Learn
• Model retirement schedule - Microsoft Foundry | Microsoft Learn
• Azure OpenAI in Microsoft Foundry Models working with models - Microsoft Foundry | Microsoft Learn

Thank you

hello @Alex ,

Checking in to see if the above response was helpful.

Thank you

That answer is precisely the problem: Azure’s “version-level” explanation does not resolve the inconsistency with official OpenAI timing or Microsoft’s own public statement that Foundry would continue Sora 2 support for six months. Since there is no replacement Sora version or alternative video generation model available in Foundry, a June 2 retirement means customers relying on video generation are effectively forced off the Foundry platform, not simply asked to migrate versions. Please check with the Azure OpenAI / Foundry team whether this retirement can be extended to Sept 24 so the platform aligns with OpenAI’s timeline and avoid unnecessary customer disruption.

Hello @Alex ,

We fully understand the concern here, particularly given the earlier public messaging around broader Sora 2 availability and the absence of an alternative hosted video generation model within Foundry at this time.

After checking internally, the retirement timeline currently shown in Azure AI Foundry for the deployed Sora 2 model version remains the effective and enforced lifecycle date for that deployment. At this time, we do not have the ability to extend or override the retirement window for that specific model version beyond the published retirement date shown in the portal.

We also want to acknowledge the confusion created by the difference between broader public communications and the version-specific lifecycle enforcement within Azure AI Foundry. The public references to continued Sora 2 support describe overall service availability at a broader platform level, whereas Azure AI Foundry applies retirement policy at the individual deployed model-version level. We understand that, in practice, this distinction is difficult when no successor video model is currently available within Foundry, and we recognize the operational impact this creates. While we cannot commit to a retirement extension or timeline change, the feedback is considered

For planning purposes, we recommend continuing to treat the retirement date currently displayed in the Azure AI Foundry portal as the authoritative date for this deployment version, since inference requests are expected to stop functioning after retirement.

We thank you for highlighting this scenario in detail. The feedback helps us to further improve and ensure better alignment in platform communications and lifecycle visibility.

Thank you