OpenAI Puts GPT-5-Level Reasoning Into Voice Model
OpenAI’s new voice models cut live translation cost to $0.034 per minute.
“AI Disruption” Publication 9800 Subscriptions 20% Discount Offer Link.
OpenAI has launched three new real-time voice models that not only integrate GPT-5-level reasoning capabilities but also deliver a powerful blow to the simultaneous interpretation industry:
Simultaneous interpretation that closely follows the speaker’s rhythm is now just $0.034 per minute.
The three models are GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. They combine end-to-end reasoning speech, simultaneous translation, and streaming transcription into a single API.
The results are mind-blowing.
OpenAI employee Jason Liu spoke English into the microphone, and GPT-Realtime-Translate instantly converted it into Japanese in real time.
The entire process is fully streaming — translation begins immediately without waiting for the speaker to finish a complete sentence.
Netizen Claire Vo used ChatPRD combined with GPT-Realtime-2. She spoke into the microphone: “Help me write a product requirements document.”
