What OpenAI announced
OpenAI introduced GPT-Live on July 8, 2026, beginning a global rollout of GPT-Live-1 and GPT-Live-1 mini in ChatGPT Voice. The company described a system that can listen and speak simultaneously, allowing interruptions and pauses without requiring a rigid exchange of turns.
For questions needing search or deeper reasoning, the voice model delegates work to another model. At launch, OpenAI identified that model as GPT-5.5. This separates the conversational interface from the system doing more complex processing.
What users should check
The announcement said GPT-Live-1 would serve Go, Plus and Pro users and the mini version would serve Free users. These are launch details, not a guarantee of today’s account availability.
OpenAI acknowledged gaps in fluency for some languages and said video and screen sharing were unsupported at launch. Its July 31 update added SynthID watermarking to supported generated audio. More natural speech does not establish factual accuracy: verify consequential answers and test pronunciation, interruptions and background noise in the setting where you intend to use it.
