Azure sora-2: July 15, then October 15, and no successor listed
Azure listed sora-2 2025-10-06 to retire 2026-07-15 and sora-2 2025-12-08 to retire 2026-10-15. Only the first has a replacement. How to read a version chain.

On the Microsoft Foundry retirement schedule, sora-2 2025-10-06 is dated to retire 2026-07-15 and points to sora-2 2025-12-08 as its replacement, while 2025-12-08 retires 2026-10-15 and lists none. The upgrade bought about three months, not a long runway. A version pin is a retirement date, so read the date before you adopt a Preview model.
The two rows (read 2026-10-04)
Both rows are labelled Preview. The page's own date is 2026-09-21.
| Version | Lifecycle | Retirement | Replacement | Days from 2026-10-04 |
|---|---|---|---|---|
| 2025-10-06 | Preview | 2026-07-15 | sora-2 (2025-12-08) | already past |
| 2025-12-08 | Preview | 2026-10-15 | None | 11 |
What the chain teaches
When a replacement is another version of the same model, the clock restarts only a little. A replacement of None means your code needs a different model family, not a different version string.
- Keep the model id in config, not in code.
- Keep a second model tested and warm.
- Track the retirement date in your calendar the day you adopt a Preview model.
The same rule applied to a Sume catalog
Sume's catalog uses bare ids and GET /v1/videos/models is the live list. Instead of pinning one id, read the list when you start and pick by supported_durations, supported_resolutions and generate_audio. A model that disappears from the list is then a config change, not an outage.
curl "https://api.sume.com/v1/videos/models" \
-H "Authorization: Bearer $SUME_API_KEY"What OpenAI did
OpenAI's deprecations page lists the Videos API and the Sora ids as removed on 2026-09-24 with no recommended replacement, so the Azure date is the only runway that remained after that.
Sources
Related posts
More in Models
- ChatGPT picks Flare or Sunburst for you; the API does not
In ChatGPT, Flare handles standard use and Sunburst takes complex prompts. In the API you choose. How to build that rule on Sume, where sume/auto uses Flare.
- Choose an AI video model by the asset you already have: Sume ids
Photo, two frames, product shots, a voice track, a driving clip or a finished video: which Sume video id fits each starting asset, from the catalog rows.
- Swap the gift in a Christmas clip with Gemini Omni Flash edit
Gemini Omni Flash 1.1 on Sume edits a 3 to 10 second clip by prompt, so one approved Christmas ad can become three gift variants without a reshoot.
- Does Lyria 3.5 audio carry a SynthID watermark?
Yes: Google's Gemini API music docs say Lyria audio carries a SynthID watermark. The Sume docs I read do not describe watermarking either way.
Written by Sume