Perplexity has upgraded voice mode in its Mac app to run on OpenAI’s gpt-live-1 voice model, so the assistant can now listen while it speaks. Announced on October 28, 2026, through Perplexity’s Computer account on X, the update lets users work across multiple Computer sessions at once, start new tasks, and create and manage Projects in a single voice session. It is available now on the Mac app and on the web for all Pro and Max subscribers, while free users are left out of the upgrade.
Listening While Speaking Is The Point
The key capability of gpt-live-1 is full-duplex conversation: the model keeps taking in what the user says while it generates its own answer, deciding many times per second whether to talk or wait. That is what makes a voice session feel like a conversation rather than a walkie-talkie exchange, where each side must finish before the other begins. For a tool built around doing work, not just chatting, the difference is practical: a user can interrupt, correct course, or add detail mid-answer without waiting for the assistant to stop talking.
Perplexity’s announcement frames the upgrade around work, not small talk. A single voice session can now span multiple Computer sessions, launch new tasks, and manage Projects, the folders where Perplexity organises ongoing research. That positions voice mode as a control surface for the agent, not a separate chatbot: speak a task, watch Computer carry it out across apps and files, and steer it by voice as it goes.
OpenAI introduced GPT-Live, the engine behind gpt-live-1, in July as the new engine for ChatGPT Voice, and it began rolling out a newer gpt-live-1.1 in ChatGPT Voice in mid-October. Perplexity’s post names gpt-live-1, the version OpenAI released through its API in September, rather than the newer point release. Perplexity has also not said which backend model its voice mode uses for questions that need web search or deeper reasoning, a detail worth noting before assuming parity with ChatGPT Voice itself.
Why A Rival’s Voice Model Powers Perplexity
The most interesting part of this story is who supplies the voice. Perplexity, an answer engine built to challenge Google, is now running its voice mode on a model from OpenAI, the company behind ChatGPT. That is not a contradiction so much as a statement about how the AI market now works: even direct competitors buy each other’s best components where it makes sense, and Perplexity’s bet is that its product is the agent and the search experience around the model, not the voice model itself.
It also sharpens the comparison users will make. ChatGPT Voice and Perplexity voice mode now share a voice engine, so the difference between them becomes what each assistant can actually do: ChatGPT’s breadth against Perplexity’s cited answers and its Computer agent, which works across local files and Mac apps. Perplexity recently extended that agent in the other direction too, letting Computer run fully local on a Mac, so voice becomes the hands-free front end for an agent that already lives close to the user’s files.
Who Gets It, And What It Costs
Availability is the clearest limit. The upgraded voice mode is open to Pro and Max subscribers only, which means the free tier keeps whatever voice mode it had before, without the new engine. Pro costs 20 dollars a month and Max costs 200 dollars a month, under the pricing ladder Perplexity already publishes, so the voice upgrade is effectively a paid-tier perk rather than a product-wide change. Users on the web get it too, not just the Mac app, which matters for anyone who switches between machines during the day.
Language coverage is an open question. OpenAI says it optimised GPT-Live for the most popular languages in ChatGPT, and that other languages may come with a non-native accent or gaps in fluency. Whether Perplexity’s voice mode covers the same languages is not stated in its announcement, so non-English users should test it before relying on it for daily work.
Why It Matters
Voice mode matters because it changes who can use an agent and when. Typing a complex task into a chat box is work; speaking it while doing something else is not, and a voice mode that listens while it talks removes the friction that has kept voice assistants feeling like demos. For Perplexity, the upgrade also signals a strategy: win on the agent and the answers, and rent the best voice engine available, even if it comes from a rival. The risk is dependence: if OpenAI changes pricing, access or the model itself, Perplexity’s voice experience changes with it, and subscribers have no control over that bargain.
Background
Perplexity began as an answer engine built around cited search, then expanded into an agent platform with Computer, a browser in Comet, and paid tiers up to Max for heavy professional use. Voice mode is the latest surface for that agent, and the gpt-live-1 upgrade brings it the full-duplex conversation that has become the standard for serious voice assistants.
Source: 9to5Mac report
0 Comments