gpt-4o-mini-transcribe-2025-12-15 to gpt-live-transcribe
Some values not published Checked 2026-10-24.
Price per 1M tokens
| Field | Old | New | Change |
|---|---|---|---|
| Input | Not published | Not published | Not published |
| Output | Not published | Not published | Not published |
| Cached input | Not published | Not published | Not published |
| Example monthly cost | Not published | Not published | Example only |
Context and max output
| Limit | Old | New | Change |
|---|---|---|---|
| Context window | 16,000 | Not published | Not published |
| Max output | 2,000 | Not published | Not published |
Capabilities that differ
No differing capability values are published for both models.
Endpoint changes
- File transcription sends the audio file to POST /v1/audio/transcriptions with client.audio.transcriptions.create. The speech-to-text guide says to start with gpt-transcribe for recorded speech. Source
- gpt-live-transcribe is a realtime transcription session (session type transcription, WebSocket or WebRTC), not that file endpoint. The realtime transcription guide says gpt-live-transcribe does not return word-level timestamps, speaker labels, or transcription confidence scores. Source
Checklist
- Read the shutdown date for gpt-4o-mini-transcribe-2025-12-15 on the linked official page.
- Point your requests at gpt-live-transcribe.
- Re-check price, context, and parameters on the provider pages linked here.
- Ship the change before the published shutdown date.