Perplexity Lets Users Run AI Workflows Entirely on Their Mac

Perplexity has introduced a new feature that allows users to run its Computer AI platform entirely on their own Mac, reducing reliance on cloud computing and giving users greater control over their data and AI spending.

The update builds on Hybrid Compute, which Perplexity introduced in early September. That feature allowed users to divide AI tasks between advanced cloud-based models and local models running on their own computers. The latest release takes the approach further by bringing the entire Perplexity Computer experience—including its underlying execution framework and orchestrator—onto the user’s device.

To use the fully local version, users need a recent Mac powered by Apple Silicon with at least 36GB of RAM.

Perplexity Computer Now Runs Locally on Mac

Running Perplexity Computer locally could appeal to users who want to keep sensitive information on their devices or reduce the cost of using cloud-based AI models.

For example, users can give Computer access to their Notes and Mail apps and ask it to create a prioritized to-do list based on their messages. With local processing, the workflow can be completed without sending email data to the cloud.

Perplexity says the move reflects the growing ability of personal computers to handle more demanding AI workloads.

“As more demanding workflows become practical to run locally, the machines people already own will take on a larger share of their AI work,” the company said. “This gives users more control over sensitive data and credit spend.”

The feature could be particularly useful for people who regularly rely on AI assistants to organize information, manage tasks, and work with personal data.

Qwen 3.6 and PPLX Pro Power Local AI Workflows

Perplexity Computer currently offers two models for local use: Qwen 3.6 and PPLX Pro.

PPLX Pro is a Perplexity-developed version of Qwen 3.6 that has undergone additional post-training to improve its performance on Computer-related tasks.

Users can also authorize Perplexity Computer to call on more than 15 advanced AI models when a task would benefit from additional capabilities. This provides a way to combine local processing with more powerful cloud-based models when needed.

However, using those external models means that some work may still be processed in the cloud. Users seeking a fully local workflow should therefore pay attention to when cloud-model access is enabled.

Security Remains a Key Part of the System

Perplexity says any code or tool execution required by a task takes place in an isolated sandbox environment. This is designed to separate task execution from the rest of the system and help limit potential security risks.

The company added that the same security protocols are maintained in the cloud-based version of Perplexity Computer.

Local processing can reduce the need to send sensitive information to external services, but it does not automatically eliminate every privacy or security risk. Users should still review app permissions and understand when a workflow may use cloud-based models.

How to Run Perplexity Computer Locally

The local version of Perplexity Computer is available to Perplexity Pro, Max, and Enterprise subscribers.

To get started, users need to download and install the latest version of the Perplexity Mac app. They can then follow the instructions within the app to download and install a supported local AI model.

Before installing, users should confirm that their Mac meets the hardware requirements, including an Apple Silicon processor and at least 36GB of RAM.

A Step Toward More Private, On-Device AI

Perplexity’s latest update reflects a broader shift toward running AI workloads directly on personal computers rather than relying entirely on remote servers.

By supporting local models while retaining the option to use advanced cloud systems, Perplexity Computer gives users more flexibility over how their AI tasks are handled. For people working with sensitive information or trying to manage usage-based AI costs, that flexibility could be an important advantage.

The key distinction is that local execution offers greater control, while cloud-model access remains available for tasks that need additional capabilities. How users balance those options will determine how much of their work stays on their Mac.