If you’ve ever chatted with an AI model before, you’ve probably noticed that it doesn’t quite feel like a real conversation. More often than not, it comes across as a back-and-forth Q&A, with the AI mostly agreeing with whatever you say. OpenAI is now looking to change that with its new GPT-Live model.
According to the official article, GPT-Live is built on a “full-duplex architecture”, allowing it to listen and speak at the same time. The result is a more natural conversation, with the AI responding faster instead of waiting for you to finish speaking. It can even acknowledge what you’re saying with phrases like “mhmm” or “yeah”, making the interaction feel more lifelike
The company says GPT-Live currently runs on its GPT-5.5 model. However, OpenAI also noted that it will adopt more powerful models, such as GPT-5.6, as they become available.
Furthermore, OpenAI says the system uses “a new ChatGPT Voice experience that is more intelligent and natural to use”. The company added that, over time, it aims to “unlock the ability to use voice for more complex, longer-running, and more agentic work” as the technology continues to evolve.
Starting today, OpenAI is rolling out two variants of GPT-Live globally: GPT-Live-1 and GPT-Live-1 mini. The former will become the default model for Go, Plus, and Pro users, while the latter will be available to Free users. The feature is rolling out across Android, iOS, and desktop browsers.
That said, after downloading the ChatGPT app, we found that GPT-Live is not yet available in Malaysia. It is likely being rolled out in phases, although OpenAI has not provided a timeline for its availability here. The company also said it plans to bring the models to its API in the near future.