Perplexity Computer Adds Fully Local AI Execution on Apple Silicon Macs

Perplexity announced that its Computer application now supports fully local execution on Apple silicon Macs. This update allows Pro, Max, and Enterprise subscribers to run AI workflows entirely on-device, keeping personal data private. According to MacRumors, the feature is available immediately for eligible users.

The local implementation utilizes Qwen 3.6 or PPLX Pro, a post-trained variant of Qwen, within the Perplexity Mac app. Perplexity states that the AI model, task planning, reasoning, and file processing all occur on the device. Code and tool execution operate within isolated sandbox environments to maintain security boundaries.

Technical Requirements and Architecture

Perplexity recommends a minimum of 36GB of memory for running the Computer app fully locally. The company highlights the new Mac mini machines as ideal candidates for this workload. With this setup, the application can interact with native macOS apps such as Notes, Messages, Mail, Keynote, and Numbers. Personal data remains on the device during these operations, avoiding cloud transmission.

This development aligns with Perplexity’s earlier move to enable local workflows, as Threat Times reported on October 20, 2026. The new update expands on that capability by detailing the specific model architectures and memory thresholds required for optimal performance on Apple hardware.

Hybrid Compute Options

For users with less powerful hardware, Perplexity offers a Hybrid Compute mode. This option combines frontier models running in the cloud with local models that process information privately. Hybrid Compute functions on Macs with as little as 24GB of memory, providing an alternative for those who cannot meet the 36GB recommendation for full local execution.

When running locally, users can still authorize cloud access if a task requires current information, browser use, or cloud-connected apps. Users decide when to spend credits for these cloud interactions. Additionally, the app allows users to call on one of more than 15 frontier models for complex problems that exceed local capabilities.

What should you do?

Perplexity Computer for Mac is available now for Pro, Max, and Enterprise subscribers. Users can download the Perplexity application for Mac and select the fully local option to begin using the feature. Those with older or lower-specification devices should consider the Hybrid Compute setting to balance performance and privacy.

What to do and how to stay safe: Perplexity

• Verify your Mac’s memory capacity to determine if it meets the 36GB recommendation for fully local AI execution.
• Review your Perplexity subscription tier to ensure you have Pro, Max, or Enterprise access required for these features.
• Configure sandbox permissions for code and tool execution to maintain isolation from your primary operating system files.
• Monitor credit usage if you enable cloud authorization for tasks requiring real-time data or browser integration.

Step-by-step guide: Cloud Misconfiguration Response: Containment, Recovery and Prevention Steps

General security guidance from the Threat Times newsroom. It is not confirmed advice from the organisations named in this story.

Frequently asked questions

What AI models run locally in the updated Perplexity Computer app?

The local version runs Qwen 3.6 or PPLX Pro, which is a post-trained version of Qwen, directly on the Mac device.

How much memory is required to run Perplexity Computer fully locally?

Perplexity recommends at least 36GB of memory for fully local operation, though Hybrid Compute works with as little as 24GB.

Can the local version access cloud services or current information?

Yes, users can authorize cloud access and spend credits when tasks require current information, browser use, or cloud-connected apps.