Azure OpenAI Sora (model sora v2025-05-02) retiring Feb 28, 2026 — no successor available (Sora 2 not visible) + video content download endpoint returns 404

Question

Severity / Impact

Severity: High (Production-impacting / service disruption imminent) Impact date: 28 Feb 2026 (retirement date shown in portal)

---

Summary

We have an Azure AI Foundry deployment of the OpenAI Sora video generation model:

Model: sora

Version: 2025-05-02

Lifecycle: Preview

Retirement date displayed in portal: Feb 28, 2026

Deployment name: hrvideo

Deployment type: Standard

Region: East US 2 and Sweden Central (tested)

The deployment has Version upgrade policy: “Once a new default version is available”, but there is no newer default version available to upgrade to, and Sora 2 (sora-2) is not available in Model Catalog for our tenant (even after creating a new Foundry resource in Sweden Central).

Additionally, our application can create jobs and poll them successfully, and the job reaches status = succeeded with valid generations[].id values — however we cannot download the MP4 via REST because the content retrieval endpoint returns HTTP 404 Resource not found.

We need guidance and enablement to prevent service outage on the retirement date.

---

What we need (requested action)

Please provide one of the following resolutions:

Enable Sora 2 (sora-2) availability for our tenant/subscription so we can deploy a successor model before the Sora v1 retirement date; or

Provide the official successor model/version that replaces sora 2025-05-02 in our tenant/region and ensure our deployment auto-upgrades prior to retirement; or

Enable / confirm the supported REST endpoint for downloading generated video content for Sora v1 (or confirm it is not supported), and provide the supported approach to retrieve MP4 assets programmatically.

---

Reproduction details

Create job (works): POST https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/jobs?api-version=preview Body example:

JSON

{

"model": "hrvideo",

"prompt": "A video of a cat",

"height": "1280",

"width": "720",

"n_seconds": "8",

"n_variants": "1"

}

Show more lines

Response returns id: task_... and status progresses to succeeded.

Poll job (works): GET https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/jobs/{task_id}?api-version=preview Response includes generations: [ { id: gen_... }, ... ]

Download content (fails): Attempted:

GET https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/{gen_id}/content/video?api-version=preview

GET https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/{gen_id}/content?api-version=preview

Both return: HTTP 404 Resource not found

---

Business requirement

We are building an internal workflow that generates short recruitment marketing videos programmatically. The model retirement on Feb 28, 2026 will cause an outage unless we can migrate to a supported successor model / endpoint.

---

References

Azure OpenAI model retirement lifecycle notes: Preview models have short lifecycles and retirement means the deployment becomes unavailable.

Severity / Impact

Severity: High (Production-impacting / service disruption imminent)
Impact date: 28 Feb 2026 (retirement date shown in portal)

---

Summary

We have an Azure AI Foundry deployment of the OpenAI Sora video generation model:

Model: sora

Version: 2025-05-02

Lifecycle: Preview

Retirement date displayed in portal: Feb 28, 2026

Deployment name: hrvideo

Deployment type: Standard

Region: East US 2 and Sweden Central (tested)

The deployment has Version upgrade policy: “Once a new default version is available”, but there is no newer default version available to upgrade to, and Sora 2 (sora-2) is not available in Model Catalog for our tenant (even after creating a new Foundry resource in Sweden Central).

Additionally, our application can create jobs and poll them successfully, and the job reaches status = succeeded with valid generations[].id values — however we cannot download the MP4 via REST because the content retrieval endpoint returns HTTP 404 Resource not found.

We need guidance and enablement to prevent service outage on the retirement date.

---

What we need (requested action)

Please provide one of the following resolutions:

Enable Sora 2 (sora-2) availability for our tenant/subscription so we can deploy a successor model before the Sora v1 retirement date; or

Provide the official successor model/version that replaces sora 2025-05-02 in our tenant/region and ensure our deployment auto-upgrades prior to retirement; or

Enable / confirm the supported REST endpoint for downloading generated video content for Sora v1 (or confirm it is not supported), and provide the supported approach to retrieve MP4 assets programmatically.

---

Reproduction details

Create job (works):
POST https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/jobs?api-version=preview
Body example:

JSON

{

"model": "hrvideo",

"prompt": "A video of a cat",

"height": "1280",

"width": "720",

"n_seconds": "8",

"n_variants": "1"

}

Show more lines

Response returns id: task_... and status progresses to succeeded.

Poll job (works):
GET https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/jobs/{task_id}?api-version=preview
Response includes generations: [ { id: gen_... }, ... ]

Download content (fails):
Attempted:

GET https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/{gen_id}/content/video?api-version=preview

GET https://<resource>.cognitiveservices.azure.com/openai/v1/video/generations/{gen_id}/content?api-version=preview

Both return:
HTTP 404 Resource not found

---

Business requirement

We are building an internal workflow that generates short recruitment marketing videos programmatically. The model retirement on Feb 28, 2026 will cause an outage unless we can migrate to a supported successor model / endpoint.

---

References

Azure OpenAI model retirement lifecycle notes: Preview models have short lifecycles and retirement means the deployment becomes unavailable.

Answer 1

Hi Dave Hughes,

Thanks for the detailed report. this is a valid concern and not caused by a configuration or deployment mistake on your side.

What’s happening

• The Sora model sora‑v2025‑05‑02 is a preview model and is reaching its scheduled retirement, as shown in the Azure portal and model lifecycle documentation.
• When a preview model is retired:
  • Existing deployments cannot be extended
  • Auto‑upgrade does not occur unless a successor model is explicitly made available for the tenant/region [learn.microsoft.com]
  • Once retired, requests may return 404 or error responses, including content retrieval endpoints

Why there’s no upgrade path visible

• Sora 2 (sora‑2) availability is tenant‑ and region‑controlled, and it’s expected that some subscriptions will not yet see a successor model in the Model Catalog.
• The “upgrade when a new default version is available” policy does not override model retirement if no successor is published for that tenant/region.

Troubleshooting & validation steps

To rule out any local issues and confirm this is lifecycle‑related:

1. Confirm lifecycle status
   • Azure AI Foundry → Model Catalog → Sora
   • Verify that sora‑v2025‑05‑02 is marked Retiring or Preview with a retirement date
2. Check Azure Resource Health
   • Resource Health surfaces model retirement notifications for active deployments
   • Notifications are sent to subscription owners and readers
3. Validate API behavior
   • Job creation and polling succeeding while MP4 download returns 404 is consistent with a retiring preview model whose content endpoints are no longer guaranteed.
4. Verify Model Catalog
   • Confirm whether sora‑2 or another Sora successor is visible for your tenant/region
   • If not visible, this is expected behavior at this stage and not an access misconfiguration

Please let me know if there are any remaining questions or additional details, I can help with, I’ll be glad to provide further clarification or guidance.

If this answers your query, please do click Accept Answer and Yes for was this answer helpful.

Thankyou!