An Azure service that provides access to OpenAI’s GPT-3 models with enterprise capabilities.
New announced speech-to-text models for Realtime
OpenAI has announced new speech-to-text models, gpt-4o-transcribe and gpt-4o-mini-transcribe. These models are also compatible with the Realtime API. However, in Azure OpenAI Services, only the whisper-1 model is currently available. Will these new models be added to Azure OpenAI Services soon?
Azure OpenAI in Foundry Models
-
Manas R Mohanty • 17,270 Reputation points • Moderator
2025-03-21T14:03:43.5333333+00:00 Hello Atakan K
Yes, gpt-4o-transcribe and gpt-4o-mini-transcribe are not available yet in Azure OpenAI.
Microsoft Azure continuously works on bringing new OpenAI model. Once added, it will be reflected in the What's New section. But there is no blog published on it yet. We recommend regularly checking the latest updates to stay informed about new releases and feature enhancements.
Feel free to post in feedback forum to bring it to product team attention.
Thank you.
-
-
-
Jurriaan • 35 Reputation points
2025-04-07T08:32:54.53+00:00 Hello Manas,
Is there an update on a timeline? Are these APIs being implemented already or are they delayed?
Thanks!
-
Manas R Mohanty • 17,270 Reputation points • Moderator
2025-04-07T08:41:27.1433333+00:00 Hi Jurriaan
There is no ETA or announcement yet from product team. Please post in Feedback forum to request the model availability.
Please check out the Realtime audios models instead
Thank you for understanding.
-
Jurriaan • 35 Reputation points
2025-04-07T09:32:17.7233333+00:00 The realtime audio API is a multitude more expensive. We currently use the whisper API and these new transcribe APIs are less expensive and more accurate compared to whisper.
-
Atakan Kurses • 10 Reputation points
2025-04-11T07:09:48.7366667+00:00 The Transcribe APIs actually seem more accurate. I also posted in the feedback forum, and I hope these models will be added soon.
-
Manas R Mohanty • 17,270 Reputation points • Moderator
2025-04-16T15:37:48.1866667+00:00 Hi Atakan K
Wanted to update that gpt-4o-transcribe and gpt-4o-mini-transcribe are available for Global Deployment now (Added in April 2025)
Reference
whats-new#gpt-4o-audio-models-released
Thank you.
-
Atakan Kurses • 10 Reputation points
2025-04-17T12:39:24.69+00:00 Hello,
It seems great, but when I use it in realtime services it still gives this error:
Invalid value: 'gpt-4o-mini-transcribe'. Supported values are: 'whisper-1'. -
Manas R Mohanty • 17,270 Reputation points • Moderator
2025-04-17T13:38:14.54+00:00 Hi Atakan K
I am able to create deployment and test them in audio playground but don't see option to view code.
Thank you.
-
Atakan Kurses • 10 Reputation points
2025-04-18T07:20:19.7833333+00:00 I think it's not the same case. I'm trying on Realtime API deployments.
Based on this OpenAI event docs: https://platform.openai.com/docs/api-reference/realtime-client-events/session/update
I need to set input_audio_transcription model as transcribe models. It is supported in OpenAI services. -
Manas R Mohanty • 17,270 Reputation points • Moderator
2025-04-18T08:04:24.2633333+00:00 Ok. Got it. Thank you for the update.
Guess we have to wait for its availability in Azure OpenAI side then.
Sign in to comment