GPT-Transcribe
OpenAI's recommended speech-to-text model for recorded audio and batch transcription
Context window
25 MB audio files
Input / 1M tokens
$0.0045 per audio minute
Output / 1M tokens
Not announced
Provider
OpenAI
Billed per minute of audio, not per token: $0.0045 per minute. The streaming GPT-Live-Transcribe costs $0.017 per minute of realtime audio. OpenAI has deprecated whisper-1, gpt-4o-transcribe, gpt-4o-mini-transcribe and gpt-4o-transcribe-diarize, which shut down on February 26, 2027. · Data verified 2026-11-11
GPT-Transcribe (July 28, 2026) is OpenAI's recommended model for transcribing recorded speech in its original language. It handles completed audio files, streamed file transcripts and committed turns in Realtime transcription sessions, and accepts context text, keyword hints and multiple language hints to help with names and technical terms. It costs $0.0045 per audio minute. OpenAI launched it with GPT-Live-Transcribe (gpt-live-transcribe), a low-latency streaming model for live captions at $0.017 per minute; according to GIGAZINE, OpenAI's charts show lower error rates for GPT-Transcribe than whisper-1 and for GPT-Live-Transcribe than gpt-realtime-whisper. OpenAI has deprecated whisper-1 and the GPT-4o transcribe models, which shut down on February 26, 2027.
How to access it
OpenAI API: the Transcription endpoint (v1/audio/transcriptions) for files up to 25 MB (mp3, mp4, mpeg, mpga, m4a, wav, webm) and Realtime transcription sessions (v1/realtime/transcription_sessions). For live captions OpenAI recommends gpt-live-transcribe instead.
Strengths
- ✓Low price at $0.0045 per audio minute
- ✓Context text, keyword hints and multiple language hints
- ✓Works for files and for committed turns in Realtime transcription sessions
- ✓Streaming transcripts for long files
Best for developers who...
When to choose it (and when not to)
Reach for GPT-Transcribe when...
- →You transcribe recorded audio files at volume
- →You need better handling of names and jargon through keyword and context hints
- →You are replacing whisper-1 or gpt-4o-transcribe before they shut down
Look elsewhere if...
- ✕You need live, low-latency captions (use GPT-Live-Transcribe)
- ✕You need timestamp_granularities, which OpenAI supports only on whisper-1
- ✕Files over 25 MB, unless you split them first
How to use it
- ›Pass product names and jargon as keyword hints
- ›Give language hints for multilingual recordings
- ›Split long recordings into chunks under 25 MB
Quickstart
Pythonfrom openai import OpenAI
client = OpenAI()
with open("recording.wav", "rb") as f:
result = client.audio.transcriptions.create(model="gpt-transcribe", file=f)
print(result.text)Install the official SDK with pip install openai. Sample from OpenAI's speech-to-text guide.
API model id: gpt-transcribe
Compare GPT-Transcribe
Compare GPT-Transcribe with any other model
Build a comparison →All model comparisons →