has anyone used the 40-realtime-cluster in production? I understand it's in beta but everything else is just way too slow with telephone communication. trying to understand the downside/risk and others experiences with it before I end up using it.

**The

4o-realtime-cluster

(OpenAI Realtime API) is currently in beta and not recommended for production use by OpenAI.** - Only OpenAI voices are supported, and Knowledge Bases are not available with this model. - While it offers much lower latency for telephone communication, you should be aware of potential instability or missing features. If you decide to proceed, monitor closely and be prepared for possible changes or issues as the feature matures. [See official documentation for details.](https://docs.vapi.ai/openai-realtime) Source: - [OpenAI Realtime API documentation](https://docs.vapi.ai/openai-realtime)

The 40-realtime-cluster in Vapi offers faster response times ideal for voice calls, but since it’s in beta, it may have occasional instability, limited support, and potential downtime, you can also share your feedback at https://roadmap.vapi.ai/

**The

4o-realtime-cluster

(OpenAI Realtime API) is in beta and not recommended for production use by OpenAI.** - It offers much lower latency for telephone communication, but you should expect possible instability, missing features (like Knowledge Bases), and only OpenAI voices are supported. - If you choose to use it in production, monitor your deployment closely and be prepared for changes or issues as the feature matures. For more details, see the [official documentation](https://docs.vapi.ai/openai-realtime). Source: - [OpenAI Realtime API documentation](https://docs.vapi.ai/openai-realtime)

ah, that's fair. Is there any word down the grape-vine of when it'll be moved out of beta?

There is currently no official word or timeline on when the

4o-realtime-cluster

(OpenAI Realtime API) will move out of beta. The documentation and changelogs do not mention any planned release date or general availability. For the latest updates, it's best to monitor the [official changelog](https://docs.vapi.ai/changelog) or [OpenAI Realtime API documentation](https://docs.vapi.ai/openai-realtime). Sources: - [Vapi Changelog](https://docs.vapi.ai/changelog) - [OpenAI Realtime API documentation](https://docs.vapi.ai/openai-realtime)

Currently, I don't know, but your feedback will be appreciated..

The

4o-realtime-cluster

is currently in beta and not recommended for production use by OpenAI. While it offers much lower latency for telephone communication—which can help with speed issues—there are some important downsides to consider: only OpenAI voices are supported, Knowledge Bases aren't available, and you may encounter instability, missing features, limited support, or even downtime. If you decide to use it in production, it's a good idea to monitor things closely and be prepared for possible changes or issues as the feature continues to develop. There isn't detailed information available about specific user experiences or how severe the issues might be in real-world production, but these are the main risks highlighted so far. Source: - [OpenAI Realtime API documentation](https://docs.vapi.ai/openai-realtime)

If you see good results with the model and would like to see it in VAPI, please share your feedback

We have been using the Realtime API with alot of success. It does however not have knowledge base functionality, and manual function calling to retreive data also does not work. So all knowledge for the AI needs to be in the prompt, which get expensive. Also, it has been down for the past couple of days for the first time since we started using it, and this has lead to our products breaking. So we need that addresses ASAP by Vapi.