Introducing GPT-Live

Source: OpenAI Date: 2026-07-08 URL: https://openai.com/index/introducing-gpt-live

Summary

OpenAI launches GPT-Live-1 and GPT-Live-1 mini, a full-duplex voice model family replacing Advanced Voice Mode as the default in ChatGPT (Go/Plus/Pro get the full model, Free gets mini) across iOS, Android, and web. The architecture listens and speaks simultaneously, handles interruptions and thinking pauses, and delegates to GPT-5.5 in the background for anything requiring search or deep reasoning — keeping the conversation flowing while heavier work completes asynchronously. OpenAI reports strong preference over the prior voice mode in matched conversation tests.

Implications

Feeds the closed-frontier clock as a product-layer release riding on GPT-5.5 rather than a new base model — the “delegate to a bigger model in the background, keep the surface responsive” pattern is itself notable UX architecture. Also touches agent-layer convergence: full-duplex, interruption-tolerant voice with async task delegation is the same shape of problem harnesses face with long-running agent tasks (keep the interface live while the real work happens elsewhere), so the pattern is worth tracking for crossover into agentic tooling, not just consumer voice chat.