OpenAI launched ChatGPT Work on July 9, an agent that handles research, document creation, spreadsheets, and presentations end-to-end. Users describe a task, and Work researches it, pulls data, formats results, and hands back finished work.

The same day, OpenAI released GPT-Live, a full-duplex voice model that listens and speaks simultaneously. Conversations feel less like speaking to a chatbot and more like talking to a person. The model interjects with “mhmm” or “yeah” to signal engagement, allows natural back-and-forth, and stays quiet when you’re thinking.

What Work Agent Does

ChatGPT Work connects to your files and apps. Tell it to “build a sales report for Q3,” and it researches your data, formats findings, creates charts, and returns a polished document. No copying and pasting. No manual formatting.

It produces Word docs, spreadsheets, and presentations directly. The agent works across devices—start on desktop, continue on mobile, pick up where you left off.

GPT-Live Voice Features

Voice wasn’t just made faster. GPT-Live was rebuilt to handle simultaneous listening and speaking. That’s the technical detail that matters: full-duplex architecture, not just speed improvements.

Paid users get GPT-Live-1. Free users get GPT-Live-1 mini. Both can now listen and respond at the same time, scrubbing the awkward pause that plagued earlier voice chat.

Desktop App Updates

OpenAI redesigned its desktop app to make switching between Chat and Work easier. Conversations now appear together in Recents, where you can sort, filter, and pin them. One unified workspace instead of separate tabs.

Group chats are being phased out. Starting July 9, new group chats can’t be created on web, iOS, and Android. Existing groups remain, but this signals OpenAI’s bet is on 1-on-1 and agent-based workflows.

Work and voice are the two features that matter this month: the first outsources thinking, the second makes the interaction feel human.