OpenAI will retire ChatGPT’s Advanced and Standard voice on December 11

ChatGPT users who still prefer Advanced or Standard voice have until December 11, 2026, before those choices disappear from settings. OpenAI announced the retirement on November 10 and will move affected users to GPT-Live. Their saved conversations will remain, but the voice model used to continue those conversations will change.

The deadline changes the picker, not the history

OpenAI is notifying users starting November 10. Advanced and Standard voice remain available during the transition, so the announcement does not immediately cut off either option. On December 11, both will disappear from the model picker under Settings > Voice. Accounts that previously selected one of them will default to Live voice, powered by GPT-Live.

The transition does not require users to abandon existing threads. Conversations created with Advanced or Standard voice will stay in ChatGPT history and can be continued using text or voice. For a spoken continuation after the switch, GPT-Live becomes the default rather than the model originally used in that conversation.

Voice conversations with custom GPTs will also continue using Advanced voice until its December 11 deprecation, according to the notice. That places their current voice arrangement on the same timetable as the removal of Advanced and Standard from the settings picker.

From taking turns to listening while speaking

The replacement changes how a spoken exchange works, not just which name appears in settings. OpenAI describes GPT-Live as a full-duplex model, meaning it can listen and speak at the same time. It can acknowledge a speaker with brief responses such as “mhmm” or “yeah,” handle quick exchanges, or stay quiet while someone pauses to think.

Advanced voice directly processed incoming audio and generated spoken responses, but it was turn-based. It had to wait for the user to stop speaking before answering. OpenAI characterizes that arrangement as a rigid back-and-forth, in contrast with GPT-Live’s ability to keep listening while it talks.

Standard voice used a different design: one model transcribed speech, a language model wrote an answer, and a third converted that answer into speech. OpenAI says this chain made responses slow and lost information as the conversation moved between audio and text. Those are the company’s descriptions of the older systems.

GPT-Live can also hand complex questions or tasks requiring actions to OpenAI’s text models. The company says that lets it keep a conversation going while several tasks run in the background. The voice model thus handles the live exchange while delegating work that calls for a different model.

Most users have moved, but preferences remain

OpenAI introduced GPT-Live in July. Since then, it says, voice usage has nearly doubled and virtually everyone using voice in ChatGPT now uses GPT-Live. Those adoption claims underpin its stated reason for retiring the alternatives: simplifying the voice experience and concentrating improvement work on the replacement.

The company identifies several areas it is still improving: response depth and detail, continuity across conversations, and personalization. It says lessons from Advanced and Standard voice are informing that work. These are ongoing priorities, rather than a promise that every improvement will arrive with the December retirement.

OpenAI also acknowledges that some users are accustomed to the older models’ style or way of responding, and that transitions can be challenging. Its notice points them to guidance on navigating model changes, while setting a firm end date for keeping either older voice selected.