Skip to content
Neural Daily

ToolsNews

Perplexity Computer now runs fully local on Mac

Perplexity Computer can now run entirely on an Apple silicon Mac. The model, the harness and task execution all stay on the machine. Perplexity recommends at least 36 GB of memory.

By: Mikola Máté Patrik · · 2 min read

Perplexity announcement graphic for Computer running fully local on Mac, on a dark background
Image: Perplexity · Press kit · Official announcement social image (og:image) · Source

What changed?

On October 20, Perplexity announced that its Computer agent can now run fully local on Mac (Perplexity). The model, the harness and task execution all stay on the device. With permission, Computer works with local files and Mac apps such as Notes, Messages, Mail and Numbers, without sending personal data to the cloud.

How it works

  • There are two local model options: Qwen 3.6 and PPLX Pro, a post-trained version of Qwen.
  • The Perplexity Mac app offers two modes. Hybrid Compute starts in the cloud and processes private files locally. Fully local mode starts on the Mac and keeps planning, reasoning and file processing on device.
  • When a task needs current information, browser use or cloud-connected apps, users can authorize cloud access. The local orchestrator can then call on one of more than 15 frontier models.
  • Code and tool execution run in isolated sandboxes with controlled access to files and apps.

Who can use it

Pro, Max and Enterprise subscribers, in the Perplexity Mac app on supported Apple silicon Macs. Perplexity recommends 36 GB of memory or more for local runs and points to the new Mac mini with M6 and M5 Pro. You download and set up the local model inside the app, then choose the fully local option for a task.

Why it matters

Perplexity brought Computer to local files and Mac apps in May and added Hybrid Compute in September. Now the whole task can stay on the machine. That gives two things: sensitive data never leaves the Mac, and credit spend is easier to control.

The 36 GB recommendation keeps this on higher-end machines for now. The announcement gives no numbers on how far the local model trails cloud frontier models on complex tasks. If you have avoided agents on confidential files because of the cloud, this is worth a try.

    Produced with the help of AI, reviewed by an editor. How we make our articles

    Reviewed and edited by the Neural Daily desk. About

    Related articles