OpenAI’s ChatGPT revolutionized the artificial intelligence industry with an intelligent bot that can chat with you. However, chatting with ChatGPT (as with most competitor models) doesn’t feel like a real conversation; it’s more of a Q&A session with a really knowledgeable nerd.
On Wednesday, OpenAI launched GPT-Live, a new model that has a “full-duplex architecture,” allowing it to listen and speak at the same time. This enables it to keep a far more realistic conversation, including giving you faster answers, and letting you know it’s paying attention with an occasional “mhmm.”
We haven’t been able to test it out yet, though OpenAI posted a video showing the model’s capabilities. The conversation does sound more natural, though the additional chatter probably takes some getting used to.
Mashable Light Speed
Google has a similar feature called Gemini Live, which can also keep up a longer conversation, though it doesn’t seem to be as chatty as GPT-Live.
OpenAI says GPT-Live is the smartest voice model it has launched. Currently, it uses GPT-5.5 in the background, but the company says that will be “continuously updated” as new models are released. This is notable, as OpenAI is scheduled to launch its most powerful model yet, GPT-5.6, today.
Right now, GPT-Live is being rolled out to users globally, in two flavors: GPT-Live-1 and GPT-Live-1 mini (it’s not yet available for us). OpenAI says it’ll bring these models to its API “soon.”
Disclosure: Ziff Davis, Mashable’s parent company, in April 2025 filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.
