OpenAI Launches GPT-Live: Full-Duplex Voice AI That Listens and Speaks Simultaneously
Summary: OpenAI shipped GPT-Live on July 8, a full-duplex voice model that can interrupt, respond mid-sentence, and carry on naturally — replacing the previous turn-based voice experience in ChatGPT.
Key Points
- Full-duplex architecture: unlike the old half-duplex voice mode, GPT-Live can listen and speak simultaneously, enabling natural back-and-forth similar to a phone call
- Two tiers: GPT-Live-1 (paid plans) and GPT-Live-1 mini (free plan), both rolling out globally on July 8
- For complex tasks — web searches, deep reasoning, multi-step work — GPT-Live delegates to GPT-5.5 in the background and returns the result in-conversation
- Built-in features include live translation, web search delegation, and intelligent task handoff
Why It Matters
The previous ChatGPT voice mode required a full pause between turns, making conversations feel robotic. GPT-Live removes that friction, bringing AI voice interaction closer to a natural human conversation. Combined with task delegation to GPT-5.5, it positions ChatGPT Voice as a capable real-time assistant rather than a novelty feature.
Read More
- OpenAI announcement — OpenAI
- TechCrunch coverage — TechCrunch