GPT-Realtime-2.1 and 2.1-mini Reach GA as Azure Schedules gpt-realtime-mini Retirement for July 23
Microsoft's official lifecycle doc confirms the July 7 realtime-voice GA pair — and a hard July 23 retirement for the October 2025 mini build that forces an in-window migration.
Microsoft's Azure OpenAI model retirement schedule — the authoritative lifecycle document — lists gpt-realtime-2.1 and gpt-realtime-2.1-mini as reaching general availability on July 7, 2026, with the gpt-realtime-mini build from October 6, 2025 retiring July 23, 2026.
Context
Realtime voice models power speech-to-speech applications in production — call centers, assistants, accessibility tools — where model retirements force code changes. The July 23 cutoff is a hard deprecation: after 15:59 on that date, the old build stops responding.
What changed
- GA pair (July 7): gpt-realtime-2.1 (retirement listed as 2027-06-25) and gpt-realtime-2.1-mini.
- Retirement (July 23): gpt-realtime-mini, the 2025-10-06 build — the only OpenAI realtime model with a confirmed in-window retirement on Azure's schedule.
- Upgrade content: the new mini adds reasoning and tool use at unchanged pricing, with trackers reporting a 25% or greater cut in p95 latency.
Why it matters
Deprecation discipline is now part of the frontier-access story: enterprises on the old mini build had roughly two weeks of overlap to migrate. The lifecycle doc also settles a circulating rumor — it lists gpt-5.4 as GA with retirement in March 2027, contradicting tracker claims of a "GPT-5.4 retires July 23" event; only the realtime-mini build retires on that date.
Details
The doc lists GA dates for the GPT-5.6 family (July 9) on the same schedule, making it the single most reliable source for Azure OpenAI availability and retirement timing.
Migration guidance
Applications calling the 2025-10-06 gpt-realtime-mini build should update model IDs to gpt-realtime-2.1-mini and re-test latency-sensitive paths before the cutoff; the new build adds reasoning and tool use, which can change streaming behavior even at unchanged pricing. Teams pinning deployments by build date in Azure Foundry should confirm their default-version settings, since GA defaults shift as new builds land.
Limitations and caveats
The latency and capability characterizations of the 2.1-mini upgrade come from tracker coverage rather than the lifecycle doc itself, which records dates only. Migration guides should be taken from Azure's official documentation.
Sources
*Update note: This post was last reviewed on 2026-07-22. The gpt-realtime-mini retirement takes effect July 23, 2026; applications still calling the 2025-10-06 build should migrate immediately.*
Sources
Drafted with AI assistance from source briefs; reviewed for citation completeness and label accuracy.