OpenAI launches three new real-time voice models for reasoning, translation, and transcription, included in its Realtime API (Zac Hall/9to5Mac)