OpenAI launches GPT-Live ahead of expected GPT-5.6 debut, bringing more natural voice conversations to ChatGPT

OpenAI has introduced GPT-Live, a new voice model that will power ChatGPT Voice, bringing more natural conversations, live translation, improved turn-taking and smarter responses to the AI assistant. The announcement comes at a busy time for the company, which is also expected to unveil its next-generation GPT-5.6 family of models later this week. The new lineup is expected to include flagship and lightweight variants—GPT-5.6 Sol, Terra and Luna—bringing improvements in reasoning, coding and multimodal capabilities across ChatGPT and the API.

GPT-Live is OpenAI's most advanced voice model to date and is built on a full-duplex architecture, allowing it to listen and speak simultaneously. Unlike previous voice models that alternated between listening and responding, GPT-Live can process speech in real time, making conversations feel more fluid and reducing awkward pauses or interruptions.

One of the biggest upgrades is the model's ability to better understand conversational timing. Users can pause naturally, interrupt ChatGPT mid-response or simply let it listen without it immediately trying to answer. The model is also designed to handle background noise more effectively.


OpenAI says that when users ask questions requiring web searches or more complex reasoning, GPT-Live can automatically hand off the request to one of the company's frontier reasoning models before returning the answer within the ongoing conversation. Users can also choose between Instant, Medium and High reasoning modes depending on the complexity of the task.

The launch builds on OpenAI's expanding portfolio of AI models. Alongside its GPT-series models for general-purpose language tasks, the company now offers dedicated reasoning models in its o-series, multimodal models capable of understanding text, images and audio, and specialised models for coding and agentic workflows. GPT-Live acts as the conversational front end, drawing on these more capable models whenever additional reasoning or web access is required.

The new voice experience also introduces visual response cards during conversations. These can display information such as weather updates, sports scores and stock prices without interrupting the ongoing discussion.

Another major addition is continuous live translation. Because the model can listen and speak simultaneously, conversations between people speaking different languages can continue in real time without repeatedly restarting the interaction.

According to OpenAI, more than 150 million people use Voice and Dictation features in ChatGPT every week. GPT-Live is rolling out globally starting today and will become the default voice model for ChatGPT across mobile and the web.

The full GPT-Live experience will be available by default for Go, Plus and Pro subscribers, while Free users will receive a lightweight mini version of the model. API access for developers is expected at a later date.
The launch comes as competition in AI voice assistants continues to intensify. Google has been expanding Gemini Live across Android devices, while Meta has integrated voice capabilities into Meta AI across its platforms. OpenAI's latest update reflects the industry's broader shift from text-based chatbots to AI assistants capable of holding natural conversations, reasoning across multiple tasks and eventually completing longer, agentic workflows through voice.