ST-2026-086 · API endpoints & versions
OpenAI schedules legacy audio, realtime and transcription model removals
OpenAI deprecated a set of legacy audio, realtime and transcription model families and snapshots for removal from the API on 20 January 2027.
Previous state
Applications could call the listed legacy gpt-realtime, gpt-audio, gpt-4o audio/realtime and transcription model families or snapshots.
Current state
The listed models are deprecated and scheduled for removal, with model-specific replacements including gpt-realtime-2.1, gpt-realtime-2.1-mini, gpt-audio-1.5 and newer transcription snapshots.
Affected users
Who needs to care
OpenAI API applications that reference any model or snapshot in the 20 July 2026 audio, realtime and transcription retirement table.
Required response
What to do
Inventory affected model identifiers, select the documented replacement for each one and complete compatibility and quality testing before 20 January 2027.
Evidence boundary
What the source does not prove
The notice retires listed model families and snapshots, not the audio, realtime or transcription API capabilities as a whole. Replacement identifiers differ by model and must be followed individually.
Lifecycle history
Dated event sequence
- Deprecation notified
OpenAI notified developers and published model-specific replacement mappings.
- API removal
The listed legacy families and snapshots are scheduled to become unavailable.
Evidence ledger
First-party sources
- 01OpenAI Developers — OpenAI API deprecations — legacy audio, realtime, and transcription models
Official API model deprecation and replacement table · 2026-07-20
Open official source ↗