2 ms·

We use it but limited to APIs. Their fast-think-2.0 model for voice conversations is a lot more natural than the openai realtime2.1