OpenAI has launched three new voice AI models: GPT-Realtime-2, GPT-Realtime-Translate, and GPT-Realtime-Whisper. These tools make AI better at handling real-time speech, reasoning, and tasks during live conversations.
GPT-Realtime-2 adds strong reasoning like GPT-5, supports multiple tools at once, speaks while thinking, and controls tone for more natural talk. On the Big Bench Audio test, it scored 96.6%, up from 81.4% for the previous version—a big leap in real-time smarts.
The new translator handles over 70 languages live, and the transcription model streams speech to text. Companies like Zillow, Priceline, and Deutsche Telekom are already using them for AI agents in real estate, travel booking, and customer support.
These models end clunky turn-taking in voice AI, enabling smooth workflows with tools and reasoning. This shifts AI from text chats to natural voice interactions.
SPEECH AI IS NOW AS FAST AS REAL TALK—GET READY TO DITCH TYPING FOREVER!
Sanjay Sahay
Have a nice evening.