OpenAI Transcription Models

[garn]

Has anyone been successful in getting OpenAI transcription to work using GPT4o or GPT4o Mini models?

GPT4-Turbo works just fine, but not the other options. They are included in my OpenAI's project model limits.

Thank you,

Garan

 

[nikolascx]

Hi @garn
They should work. Because before exposing them in the dropdown, we make sure that what we expose, works.

If Turbo works but 4o / 4o-mini fail the most likely causes are the following:

Your OpenAI project/org doesn’t actually have access to the audio transcribe variants
Even if you “see” the models in limits, the API call can still fail if the project is missing access to gpt-4o-mini-transcribe / gpt-4o-transcribe. (these are the actual model names).

Could also be you are using a restricted key/project.

The 3CX version is not an issue because from the moment of release, we always shipped support for all the gpt4 models. But in your reply, it is good to specify it.

Check the logs. System service logs.

 

[garn]

Hey, I will revisit at some point, check those things, and update!

I did initially check system logs and did not see anything, and also saw no activity on the OpenAI side, which could point to some of the blocks you mentioned.

 

[nikolascx]

Coincidentally, today we did some work on this area.
The Gemini 2.0 will be retired by Google this March. This means we had some work to do to update all users to the Gemini 2.5 model and make a full retest of everything. (This is coming in the next update)
Since we had a retest to make, we also made a full retest of all the models and LLM's.

I can confirm that all advertised models of OpenAI work.

Thanks so much!