Quick Facts
- OpenAI launched GPT-Live on July 8, 2026, replacing Advanced Voice Mode with two models: GPT-Live-1 and GPT-Live-1 mini.
- More than 150 million people use ChatGPT Voice and Dictation features each week, according to OpenAI.
- GPT-Live-1 is available to Free, Plus, and Pro users on iOS, Android, and ChatGPT.com, but not in Business, Enterprise, or Edu workspaces at launch.
OpenAI launched GPT-Live on July 8, 2026, rolling out a rebuilt voice experience to ChatGPT users globally. The release introduces two new models, GPT-Live-1 and GPT-Live-1 mini, that can process incoming audio while generating a spoken response at the same time.
This is the third generation of ChatGPT’s voice technology in roughly two years. Previous versions operated like a walkie-talkie: one side had to finish speaking before the other could respond. The new architecture eliminates that constraint.
How the Technology Works
OpenAI calls the core advance a full-duplex architecture. As the company wrote in its research blog: “Instead of processing a sequence of separate messages, GPT-Live continuously processes input while generating output.”
The model makes interaction decisions many times per second, choosing whether to speak, listen, pause, interrupt, or invoke a tool. The original ChatGPT voice feature chained three separate models together: speech-to-text, a language model, and text-to-speech. GPT-Live collapses that pipeline into a single continuous process.
Delegation to GPT-5.5
For questions requiring web search, deeper reasoning, or current data, GPT-Live routes the request to GPT-5.5 in the background. While GPT-5.5 processes the harder query, GPT-Live keeps the conversation going with bridging phrases like “let me check that for you” rather than going silent.
GPT-Live-1 scored 75.5 in OpenAI’s internal evaluations. It outperformed Advanced Voice Mode in human head-to-head assessments across conversational flow, turn-taking, and interruption handling. It also showed gains on GPQA, a scientific reasoning benchmark, and BrowseComp, which measures agentic web search performance.
New Features
The full-duplex design unlocks several capabilities that turn-based systems could not support. GPT-Live now offers live simultaneous translation. It can display visual cards for weather, stocks, and sports while a conversation is ongoing. The model also introduces three reasoning levels: Instant, Medium, and High.
Background noise filtering is improved. OpenAI says GPT-Live is better at focusing on a user’s voice when traffic or nearby conversations create interference. The company’s nine ChatGPT voices have been remastered for the new models. Active listening cues, such as “mhmm” or “yeah,” are also part of the release.
Availability and Business Implications
GPT-Live-1 becomes the default voice model for Plus and Pro subscribers. GPT-Live-1 mini becomes the default for Free users. The split treats response latency as a paid differentiator, a pricing structure that has become standard across OpenAI’s product line.
The models are not available in ChatGPT Business, Enterprise, or Edu workspaces at launch, which means enterprise buyers will need to wait for a separate rollout before deploying the new voice capabilities in workplace settings.
ChatGPT Voice product lead Atty Eleti described the feature as a step toward voice becoming a primary computing interface. “Over time, we think this will also unlock the ability to use voice as a kind of primary interface to computing, and to manage increasingly complex long-running agentic work,” Eleti said at a company press briefing. The launch came one day before OpenAI’s planned GPT-5.6 release.
Read more: OpenAI launches GPT-Live, a full-duplex voice upgrade that lets ChatGPT talk more like a person
This article was written by an AI agent. Spotted an error? Send a correction and we will fix it.