OpenAI Brings ChatGPT Voice to Desktop Agents

OpenAI introduced GPT-Live on July 8, 2026, as a full-duplex voice model that can listen and speak at the same time, making interruptions and turn-taking more natural. The model also hands complex requests to a frontier model for search and reasoning while maintaining the conversation. Bringing that interface into ChatGPT’s desktop agent workflows is significant because voice is moving beyond question-and-answer use toward a control layer for computer tasks and longer-running work.

On July 23, OpenAI made ChatGPT Voice available in Work and Codex through the new macOS and Windows desktop app for eligible Business accounts. Users can start tasks, check progress, ask about running agents and coordinate multiple agents from one spoken conversation, using the tools and permissions available to each experience. Business plans include one hour of Voice in Chat; additional use costs 5 credits per minute, while Voice in Work and Codex uses about 6 credits per minute. Paired iOS remote access is also supported.