OpenAI ships GPT-live, its next-generation ChatGPT voice model

GPT-Live changes how ChatGPT handles the rhythm of a spoken conversation. OpenAI's July 8, 2026 launch article explains that it processes incoming speech while generating its own audio. This lets it decide whether to respond, pause or keep listening as the conversation unfolds.

More demanding work follows a separate path. GPT-Live can delegate search or reasoning to another model while continuing the conversation; at launch, that model was GPT-5.5. The design separates maintaining a responsive spoken exchange from working through a harder question.