GPT-4o mini Transcribe
OpenAI · USA · 2025
At $0.003 per minute this is the cheapest transcription model OpenAI sells — a third below the one it officially recommends. Quietly updated in December 2025, with the original March version due to shut down in January 2027.
GPT-4o mini Transcribe launched on 20 March 2025 alongside its larger sibling as the budget half of OpenAI's first GPT-based transcription pair. It shares the same shape — audio and text in, text out, a 16,000-token context window, 2,000 output tokens, knowledge to 1 June 2024 — and the same promise of a lower word error rate and better language recognition than the original Whisper models, traded against the accuracy of the full-size version. Its price is the part worth noticing. OpenAI charges $1.25 per million audio input tokens and $5 per million text output tokens, which comes to $0.003 per minute of recording: half of GPT-4o Transcribe and whisper-1, and a third below GPT Transcribe, the model OpenAI has recommended as the default since 28 July 2026. In other words, the cheapest way to turn speech into text on this platform is not the model the platform points you to, and the gap has been there since July. The model is also better maintained than its price suggests. It carries two dated snapshots — `gpt-4o-mini-transcribe-2025-03-20` and `gpt-4o-mini-transcribe-2025-12-15` — so a customer can pin the exact version they are billed for, something GPT-4o Transcribe does not offer at all. The December 2025 snapshot is now the default. The original March 2025 snapshot was placed on OpenAI's deprecation list on 20 July 2026 and shuts down on 20 January 2027, with the December snapshot named as its replacement; the model itself carries no shutdown date. Anyone still pinned to the March version has until that date to move one line of configuration.
▸News
▸Videos
No videos yet.
▸Reviews
No reviews yet. Be the first!