OpenAI · 2026-05-07 · major
OpenAI GPT-Realtime-2 — Voice Model With GPT-5-Class Reasoning, Plus Live Translate and Whisper Streaming in the API
Three new Realtime API models: GPT-Realtime-2 brings GPT-5-class reasoning to speech-to-speech, GPT-Realtime-Translate covers 70 input languages to 13 outputs, and GPT-Realtime-Whisper streams transcription live.
OpenAI ships three new Realtime API models — a reasoning-grade voice model, live 70→13 language translation, and streaming Whisper transcription.
Key specs
| Audio input | $32 / 1M tokens |
|---|---|
| Audio output | $64 / 1M tokens |
| Cached audio input | $0.40 / 1M tokens |
| Translate pricing | $0.034 / minute |
| Whisper pricing | $0.017 / minute |
| Translate input languages | 70 |
| Translate output languages | 13 |
What is it?
OpenAI made the Realtime API generally available and added three new audio models. GPT-Realtime-2 is a speech-to-speech model with GPT-5-class reasoning, two new voices (Cedar and Marin), and improved tool calling. GPT-Realtime-Translate streams real-time translation across 70 input and 13 output languages. GPT-Realtime-Whisper streams transcription token-by-token as the speaker talks.
How does it work?
All three run inside the same Realtime API session, which now also supports remote MCP servers, image inputs, and inbound phone calls via Session Initiation Protocol. GPT-Realtime-2 keeps the conversation moving while reasoning about the user's request rather than pausing to think; the translate model targets near-zero lag during back-and-forth speech. Pricing is by audio token for the reasoning model and per-minute for translate and Whisper.
Why does it matter?
Before today, real-time voice agents had to choose between fast call-and-response or smarter offline reasoning — not both. Reasoning-in-the-loop unlocks customer service, education, events, and accessibility apps that need to understand a user, take an action, and respond inside one breath. Native SIP support means voice agents can finally pick up phones without third-party telephony glue.
Who is it for?
Voice-agent builders, contact-center teams, accessibility developers
Try it
https://developers.openai.com/api/docs/guides/realtime