OpenAI is adding three voice models to its Realtime API, giving developers tools for live reasoning, speech translation, and streaming transcription, the company said.
The first model, GPT-Realtime-2, brings GPT-5-class reasoning to live voice interactions. OpenAI says the model is designed so that voice conversations remain fluid even as it works through complex requests, manages tool calls, and adapts to interruptions in the moment. It is priced at $32 per 1 million audio input tokens — or $0.40 for cached input tokens — and $64 per 1 million audio output tokens.
Live speech translation is the focus of the second model, GPT-Realtime-Translate, which supports input from more than 70 languages and delivers output in 13, without falling behind the natural rhythm of a conversation. It is priced at $0.034 per minute.
Rounding out the trio is GPT-Realtime-Whisper, a model that converts spoken audio into text in real time, prioritizing low latency so that transcription keeps up with speakers as they talk. OpenAI cited use cases including on-screen captions and automatically generated meeting notes as examples of where the model is intended to be deployed. It is priced at $0.017 per minute.
All three models are available through OpenAI's Realtime API. Developers can test them in OpenAI's Playground.