When the new two voices from OpenAI realtime is coming?

[Mukhsin0508]

OpenAI has recently released new voices with uograded capabilities. How soon can we get them in here?

Source: https://www.linkedin.com/posts/openai-for-business_today-were-introducing-our-most-advanced-activity-7366890615617318912-rd6O?utm_source=share&utm_medium=member_ios&rcm=ACoAADdjfrsB0jkdBTz7IbsLwbAhAVMYUEoQk2E

[juliannicolas90]

+1, also asking for the models themselves!

[Mukhsin0508]

+1, also asking for the models themselves!

Can u please specify which new models do u mean? Cuz we here in /realtime api used to use gpt-realtime model or smth. I am assuming it has already been migrated to gpt5 as older models have been deprecated.

[juliannicolas90]

on Azure I still see gpt-4o-realtime-preview as the only available one, and the new one is gpt-realtime.
I think the version they have on Azure is also the 2024-12 version, and there's one from 2025-06 on OpenAI.

[Mukhsin0508]

Wow, I missed this one apparently, It'd be great if we could get it asap.

[Mukhsin0508]

https://openai.com/index/introducing-gpt-realtime/

It's great! When will we get it in Azure!?

[nstijepovic]

We are expecting the new version of GPT realtime to be available this week in AI Foundry! Keep an eye on AI Foundry!

[goldzulu]

Hi i am quite confused. Here https://learn.microsoft.com/en-us/azure/ai-foundry/openai/whats-new#august-2025 it mentioned gpt-4o realtime GA but the notes refer to GPT-Realtime features. Is the Azure OpenAI version of realtime going to be called GPT-4o Realtime or will there be as mentioned above GPT-Realtime? Will the 4o realtime deployment automatically cutover to the new version?

[zantis]

Whats interesting is the blog announcement mentions the two new voices Marin and Cedar but they are not working for me. Otherwise, using the new model gpt-realtime created through AI Foundry with Direct WebRTC is working quite well. Does seem like a big improvement.

[Mukhsin0508]

@zantis

Have you even experienced sudden "No Response" from the realtime server on production!? My users have been complaining about that sometimes the AI ain't responding back anything to them, when I check the logs, logs also cut off at that time without any error messages? I was using websockets btw.

[Mukhsin0508]

@goldzulu

Typically Azure doesn't automatically migrate deployments to new model versions to avoid breaking changes. You'll likely need to create a new deployment with the GA model once available.

Model Naming Clarification
The naming has been a bit confusing in the rollout:
Current: The model is called gpt-4o-realtime-preview
GA Version: Based on the documentation, it appears the GA version will be called gpt-realtime (not gpt-4o-realtime)

@nstijepovic can u leave a like if the naming clarification is correct please?

[Mukhsin0508]

We are expecting the new version of GPT realtime to be available this week in AI Foundry! Keep an eye on AI Foundry!

@nstijepovic Is it out already?