Error code: 404 - {‘error’: {‘message’: ‘The model gpt-4o-audio does not exist or you do not have access to it.’, ‘type’: ‘invalid_request_error’, ‘param’: None, ‘code’: ‘model_not_found’}}
With audio models list for my client as only:
gpt-audio-2025-08-28
gpt-audio
gpt-audio-mini
gpt-audio-mini-2025-10-06
gpt-audio-mini-2025-12-15
gpt-audio-1.5
But why did I get the email saying gpt-4o-audio is available till January 2027?
As part of our continuous upgrade process, we are deprecating the following models:
gpt-realtime
gpt-audio
gpt-4o-audio
gpt-4o-realtime
gpt-realtime-mini
gpt-audio-mini
gpt-4o-mini-realtime
gpt-4o-mini-audio
gpt-4o-mini-transcribe-2025-03-20.
Access to these models will be shut off on Jan 20, 2027.
I don’t understand why I don’t have access until January 2027 if I got the email.
Hi Paul, thank you. Yes, same key, with all permissions. Do you mean that if it was a new key that hadn’t accessed this legacy model previously, then it wouldn’t get access until 2027? Thanks
Are you sure you were using that name before - and not just discovering it and trying it now from an erroneous email’s made-up model name?
Announcement
Shutdown date
Model family / snapshot
Recommended replacement
Jan 20, 2027
gpt-realtime
gpt-realtime-2.1
Jan 20, 2027
gpt-audio
gpt-audio-1.5
Jan 20, 2027
gpt-4o-audio
gpt-audio-1.5
Jan 20, 2027
gpt-4o-realtime
gpt-realtime-2.1
Jan 20, 2027
gpt-realtime-mini
gpt-realtime-2.1-mini
Jan 20, 2027
gpt-audio-mini
gpt-audio-1.5
Jan 20, 2027
gpt-4o-mini-realtime
gpt-realtime-2.1-mini
Jan 20, 2027
gpt-4o-mini-audio
gpt-audio-1.5
Jan 20, 2027
gpt-4o-mini-transcribe-2025-03-20
gpt-4o-mini-transcribe-2025-12-15
It seems the model IDs listed in the deprecations page (and also emails) are rather confused, that they created some alias that doesn’t exist. Actual model names with gpt-4o-audio all have preview in the name:
Alias and where it last pointed:
gpt-4o-audio-preview ->gpt-4o-audio-preview-2025-06-03
You will note that gpt-4o-realtime-preview-2024-12-17 is listed as shutoff, but there is no mention of the same date snapshot or notice applying to gpt-4o-audio-preview-2024-12-17, an oversight where we can assume they meant to kill it off also, and did.
The models API endpoint currently returns no “gpt-4o-audio” models, preview, or otherwise. They always had “preview” in the name until the “4o” name was also removed, giving the latest/last version with some concept of date snapshot:
gpt-audio-2025-08-28 (gpt-audio)
A model you can still use today until the shutoff. This is another case where the model you can continue to use by the actual underlying model ID is not specifically listed in the deprecations page, an issue with the last big blast of models announced also.
Also: these models are for Chat Completions only. OpenAI has not kept up these audio models with the 2.0 and 2.1 versions for realtime, so likely they are also not maintaining the promise of Chat Completions being supported or they are dropping the ball on improvements for it in favor of “realtime” - an endpoint that is far less exploratory and programmable. Or for specialized “app” APIs like TTS.