OpenAI had previously rolled out ChatGPT's standard voice mode and advanced voice mode, both of which relied on a turn-taking conversation mechanism. In July 2026, OpenAI introduced GPT-Live, a next-generation voice model built on a full-duplex architecture that enables simultaneous listening and responding, and can call upon frontier text models mid-conversation to tackle complex problems. Following its release, usage of ChatGPT's voice features nearly doubled, with the majority of voice interaction users migrating to GPT-Live. OpenAI has announced that it will officially retire both the standard and advanced legacy voice modes on December 11, 2026, after which it will concentrate its resources on further developing GPT-Live. Until the retirement date, users can still access the legacy modes; afterward, the legacy modes will be removed from settings, and affected users will be automatically switched over to GPT-Live, with past voice conversation records retained and available for continued use. Real-time video, screen sharing, and voice conversations with users and custom GPTs are temporary exceptions and will continue to use the advanced voice mode until December 11. The retirement decision has drawn considerable user backlash. Some users argue that the legacy modes deliver more detailed answers and better contextual continuity, while GPT-Live has issues such as interrupting users and losing context—prompting some paying subscribers to say they are considering canceling their subscriptions. OpenAI says it is incorporating lessons from the legacy models to optimize GPT-Live's performance. Notably, OpenAI previously delayed retiring GPT-4o in response to user opposition; whether it will adjust the retirement plan for the legacy voice modes this time remains to be seen.