OpenAI brings voice-controlled agentic features to ChatGPT’s mobile app

OpenAI on Wednesday began rolling out a suite of upgrades to ChatGPT Voice, connecting its spoken interface to GPT-6 models, third-party plugins, and the company's agentic Work mode across mobile and web — a move that turns voice conversations into a launchpad for multi-step tasks like drafting documents, managing calendars, and browsing the web.

Voice Meets Agentic Work

The update links OpenAI's real-time voice model, GPT-Live, to the AI models that can operate browsers and connected apps, extending capabilities that were previously limited to the desktop app. Plus and Pro subscribers can now use voice on their phones to trigger workflows in the Work tab, including creating presentations, summarizing Slack messages, and drafting emails.techcrunch+1

In a demo for Business Insider, OpenAI's product lead for ChatGPT voice, Atty Eleti, used only speech to analyze his DoorDash spending, book a pickleball court, and reorganize his calendar. "In the future, we really imagine people using voice as the primary modality to interact with AI," Eleti said.businessinsider

ChatGPT Voice can now also be powered by OpenAI's three GPT-6 models: Astra, the most capable; and the smaller Sol and Luna, which were updated to GPT-6 the day before. Plugin support for email, calendar, and Slack Salesforce, Inc. ↗ 1.84% has also been added to voice, closing what 9to5Mac called "a big disconnect" for users who previously had to switch to text to reach connected services.9to5mac

Balancing Speech and Screen

OpenAI said voice conversations will now produce richer text output, and users can switch between voice and text mid-conversation or start a task on mobile and resume it on desktop. More than 150 million people already use ChatGPT's dictation and voice tools, according to Eleti.techcrunch+1

The design reflects a deliberate tradeoff. "We don't think of this as just pure voice in and voice out," Eleti told Business Insider. "It's about voice in and then the right output back. Sometimes it's text, sometimes it's voice, sometimes it's a mix."businessinsider

A Crowded Race for the AI Assistant

The launch arrives as personal AI assistants gain momentum across Silicon Valley, with rivals including Meta's ↗ 1.02% Muse and the startup Instinct vying for users. OpenAI has been issuing a string of product releases ahead of its DevDay event next week. Bloomberg has also reported that the company is building a smart speaker for release in 2027, giving its voice investments a longer-term hardware dimension.businessinsider

Not all features are universal. GPT-6 model access and Work capabilities depend on plan tier and workspace settings, and the rollout is gradual. Meanwhile, the growing split between ChatGPT's Chat and Work modes — with GPT-6 available only in the latter — has added a layer of complexity that some users are still parsing.techradar+1