Krux
May 9, 2026
OpenAI's New Voice Models Can Translate and Act Live
Published: May 9, 2026 at 12:14 AM
Updated: May 9, 2026 at 12:14 AM
100-word summary
OpenAI just released three real-time voice models that can listen, reason, and execute tasks while you're still talking. The new APIs handle live translation across 70 languages, streaming transcription, and tool calls that let voice agents actually do things (book appointments, place orders) instead of just answering questions. Context memory jumped four times larger, from 32,000 to 128,000 tokens. Zillow, Priceline, and Deutsche Telekom are already using them. The tech moves voice AI past the chatbot era into interfaces that feel more like talking to a capable assistant than interrogating a search box.
What happened
OpenAI just released three real-time voice models that can listen, reason, and execute tasks while you're still talking. The new APIs handle live translation across 70 languages, streaming transcription, and tool calls that let voice agents actually do things (book appointments, place orders) instead of just answering questions. Context memory jumped four times larger, from 32,000 to 128,000 tokens. Zillow, Priceline, and Deutsche Telekom are already using them.
Why it matters
The tech moves voice AI past the chatbot era into interfaces that feel more like talking to a capable assistant than interrogating a search box.