AI That Listens and Speaks at Once Mimics Natural Conversation

Thinking Machines Lab announced voice AI models on 11 May 2026 that listen and speak simultaneously, rather than waiting for turn-taking. The breakthrough, led by former OpenAI CTO Mira Murati, responds in 400 milliseconds—three times faster than current systems. Full-duplex audio processing enables natural interruptions and overlaps, moving voice assistants closer to how humans actually converse.