GPT-Live (GPT-Live-1 and GPT-Live-1 mini)

2026-07-08 · Publisher : OpenAI · Domain : Artificial Intelligence

OpenAI launches GPT-Live, a new generation of full-duplex voice models that listen and speak at the same time, with GPT-Live-1 and GPT-Live-1 mini powering ChatGPT Voice.

What changes

  • Full-duplex architecture: the model decides several times per second whether to speak, listen, wait or interrupt.
  • Delegation: requests that require search or reasoning are handed to a frontier model (GPT-5.5 at launch) while the conversation continues.
  • Two versions: GPT-Live-1 and GPT-Live-1 mini, deployed first in ChatGPT worldwide.
  • SynthID watermark on generated audio since July 31, 2026, with a public verification tool.

### Specifications - Availability at launch: ChatGPT Voice (iOS, Android, web); API announced "soon". - Context, output and knowledge cutoff: not disclosed in the post. - Modalities: audio and text. - API identifiers published later: gpt-live-1. ### Pricing Pricing not disclosed at launch ($0.05 per minute for the API, published on September 10). ### Benchmarks No numerical score published by the vendor in the announcement. ### Key takeaways The voice model moves from turn-based operation to continuous conversation, and separates conversation from deep reasoning. This is the foundation of the GPT-Live-1 API.

Links

Tags : Audio, OpenAI, Voix, Full-duplex, ChatGPT Voice