ChatGPT Voice Mode Gains Plugin Access and GPT-6, Enabling Email and Calendar Control by Voice
nashnova research
OpenAI connected ChatGPT's voice mode to plugins and the GPT-6 model family, letting users drive email, calendar, and website tasks by speaking alone — the updated app rolls out globally today, turning voice from a chat tool into an action interface.
What actually changed in this upgrade?
Voice mode now connects to plugins for the first time — previously it could only chat, not call external tools.
Three GPT-6 models run behind the scenes: Astra handles multi-step computer tasks, Sol handles reasoning, Luna is the fastest and cheapest. Real-time voice interaction still runs on the dedicated GPT-Live model up front.
This means → voice is no longer just an "input method." It is now an agent that can act — you speak, it executes.
What is "Work," and how does it pair with voice?
ChatGPT Work is an agent product OpenAI launched in July — an AI assistant that can operate a browser on its own.
With voice plugged into Work, users dictate tasks on their phone and Work opens a browser to carry them out. After the call ends, the task converts to text and keeps running.
In plain terms = you call the AI, say "cancel that subscription for me," and it goes to the webpage and clicks the buttons itself — no need to touch the screen.
Why does the launch video feature "people who don't type"?
The promo's lead characters are mostly older women: one kneading dough, one pulling clay on a wheel, one sitting in a salon — hands occupied, using voice to have AI cancel subscriptions, chase refunds, and build an online pottery shop.
This reflects OpenAI's ambition for the entry point: the target user is not a programmer but an ordinary person who has never opened a command line — yet can speak.
ChatGPT Voice product lead Atty Eleti reposted the announcement with just two words: "At long last."
How much does it cost, and what are the limits?
Free users can use plugins in voice mode; desktop users also get access to the Luna model.
Plus subscribers get 3 hours of voice time per 24-hour window; the $200/month Pro tier has no time cap. Using voice inside Work triggers two meters at once — voice time and task usage.
Sensitive actions like sending an email still require a manual on-screen tap to confirm. Voice commands cannot replace written authorization — the safeguard against a misrecognized name sending an email to the wrong person.
Who is fighting the voice-entry-point battle?
OpenAI previously disclosed that over 150 million users interact with ChatGPT via voice or dictation.
Roughly ten days before this launch, the new Siri — powered by Google Gemini — shipped with iOS 27, but it currently supports only English and remains in beta.
OpenAI's own hardware is at least a year away from market; the phone is its primary voice entry point today. This means → the core test of this entry-point race is whether a voice agent can truly replace keyboard operation in daily workflows.
市场有风险，内容仅供研究参考，不构成投资建议。