Prime Highlights :
- OpenAI launches GPT-Live-1 and GPT-Live-1 mini, replacing ChatGPT’s Advanced Voice Mode.
- The full-duplex models allow natural interruptions and live translation.
Key Facts :
- Over 150 million people already use ChatGPT’s voice and dictation features.
- GPT-Live-1 mini becomes the free default; GPT-Live-1 is limited to paid tiers.
Background :
Two new models, GPT-Live-1 and GPT-Live-1 mini, have been launched by OpenAI that make their conversation more natural and are capable of handling the process of turn-taking in conversations. These models have been created as full-duplex models. This allows users to interrupt naturally and supports features such as live translation.
The company is replacing its existing Advanced Voice Mode in ChatGPT with GPT-Live-1 mini by default. Paid tier users are going to get access to the larger GPT-Live-1 model. Earlier, the voice mode relied on three separate systems: one to transcribe speech, one to generate responses, and one to convert text back into speech.
During a press briefing, OpenAI said the new models address problems such as interrupting users mid-sentence and lacking sufficient intelligence to answer questions well. The models route queries to OpenAI’s latest text models for search, reasoning or agentic tasks while keeping the conversation flowing. The system can also stay silent for extended periods, absorbing context until needed, and can present some responses visually.
ChatGPT Voice’s product lead, Atty Eleti, said voice could eventually become a primary interface for managing complex, long-running tasks, comparing its potential to the agentic capabilities already seen in tools like Codex.
OpenAI noted that more than 150 million people currently use voice and dictation features on ChatGPT. The company clarified that the update is not intended to turn the assistant into a companion and said safeguards remain in place for age-appropriate responses and support around sensitive topics like self-harm.