OpenAI Drops Three Real-Time Voice Models Into Its API
OpenAI unleashes a trio of voice models for developers: reasoning, transcription, and translation — all in real time.
OpenAI just shipped three new real-time voice models through its API, aiming to supercharge what developers can build with voice.
The lineup: GPT-Realtime-2 brings GPT-5-level reasoning to live voice interactions. GPT-Realtime-Whisper handles transcription. GPT-Realtime-Translate does exactly what the name suggests — real-time translation.
The company says the release will "unlock a new class of voice apps for developers." That's corporate-speak, but the underlying tech is legitimately broad. We're talking voice assistants, live translation tools, transcription engines, and whatever else devs dream up with GPT-5-class intelligence baked into the audio pipeline.
Three models. Three distinct jobs. All real-time. OpenAI is clearly betting that voice is the next major API battleground — and it just armed developers with significantly sharper tools to build on.