Pro and Max subscribers can now run multiple sessions and manage tasks and projects just by talking to the AI.
Perplexity’s voice mode now runs on OpenAI’s gpt-live-1 model. With it, users can run multiple sessions at once and manage tasks and projects through conversation.
The upgrade, announced on October 28, 2026, covers both the Mac app and the web interface.
What changed in voice mode
The headline capability of gpt-live-1 is full-duplex voice. The system can keep listening while it is still speaking.
Perplexity is pairing that conversational layer with real actions. According to the research findings, subscribers can now:
- Manage several computer sessions at the same time
- Kick off new tasks mid-conversation
- Create and manage projects without touching a keyboard
Access is gated behind paid plans. Voice mode is available only to Pro and Max subscribers.
The Pro tier costs $17 per month. The Max tier costs $167 per month.
Where gpt-live-1 came from
OpenAI did not build gpt-live-1 for Perplexity. The model first reached ChatGPT Voice users on July 8, 2026.
Two months later, on September 10, 2026, OpenAI opened up an API version. Developers pay $0.05 per minute for the voice layer.
The voice update also follows another recent push from Perplexity. On October 20, 2026, the company rolled out local AI execution on supported Mac devices. That feature was aimed at offline use and privacy-focused workflows.
Why this matters for the AI search race
The second group is developers and businesses watching from the sidelines. The $0.05 per minute API price gives them a concrete benchmark for adding similar voice features.
Per-minute pricing has a particular character. It is cheap for short exchanges and can add up quickly for long, chatty sessions. Any company building on it will have to think carefully about how much talking its users actually do.
For Perplexity specifically, the economics raise a fair question. The company charges flat monthly fees while relying on a model that is priced by the minute at the API level.
