Nightly digest ·
OpenAI deprecates whisper-1 and its gpt-4o-transcribe speech-to-text models — API removal 26 February 2027
Another quiet night for fresh vendor announcements, so tonight adds one detectable archive entry: OpenAI is retiring four transcription models — whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe, and gpt-4o-transcribe-diarize — from the API on 26 February 2027, with gpt-live-transcribe and gpt-transcribe as the named replacements. Also on the board today: the GitHub Copilot MAI-Code-1-Flash deprecation reached its date.
No new upcoming sunsets crossed the monitored sources or the discovery searches tonight, and no corrections are outstanding. The one new archive item is OpenAI’s transcription-model deprecation — a companion to the legacy audio and realtime models already tracked here. Tonight’s three re-verification flags were, again, the Microsoft pages that render their dates in JavaScript (Sentinel SAP connector, Azure Automation Python/PowerShell runtimes, Exchange Web Services); each was re-confirmed correct by hand against Microsoft’s release-communications API and the page metadata — not data errors.
New (archive): OpenAI deprecates whisper-1 and its gpt-4o-transcribe models — API removal 26 February 2027
On 26 August 2026 OpenAI notified developers that four speech-to-text models are deprecated and will be removed from the API on 26 February 2027: whisper-1 (the long-standing hosted Whisper endpoint — not the open-weight Whisper you run yourself, which is unaffected), gpt-4o-transcribe, gpt-4o-mini-transcribe, and gpt-4o-transcribe-diarize (the speaker-labelling variant). The models keep working through the runway; on the date, an API call that names any of them starts failing.
Anything that pins one of these IDs breaks: a meeting or interview transcriber, a podcast or voice-note pipeline, call-center analytics, an accessibility captioning feature. OpenAI points all four at the same two replacements — gpt-live-transcribe or gpt-transcribe — so the migration is mostly a rename, but re-run your tests: a newer model can shift accuracy, latency, timestamp formatting, and (for the diarize case) speaker labelling even when the request shape is unchanged.
This entry is detectable: the lastcall CLI fingerprints the four retiring model IDs, so it flags code that still names them without tripping over the gpt-live-transcribe / gpt-transcribe replacements or the surviving gpt-4o-mini-transcribe-2025-12-15 snapshot.
Reached its date today
- 10 September: the GitHub Copilot MAI-Code-1-Flash deprecation came due — the model is removed from all Copilot experiences, replaced by MAI-Code-1.1-Flash. Marked passed on the board.
On the board
- 14 September: the Relay app and Microsoft Sentinel SAP connector agent retirements.
- 17 September: the Azure Maps Render v1 API retirement.
- 24 September: OpenAI’s Sora 2 video API shutdown.
Further out: a cluster on 30 September, Exchange Web Services in Exchange Online (1 October), the GitHub Copilot model deprecations (2 October), OpenAI’s 23 October API removals, then .NET 8 and 9 end of support (10 November) and PostgreSQL 14 end-of-life (12 November). The full board with live countdowns is on the calendar.