News in Short

• Perplexity has launched Hybrid Compute for Apple Silicon-powered Macs.
• The feature splits AI tasks between local models and cloud models.
• A local privacy gate checks information before it leaves the Mac.
• Sensitive data can stay on the device, while the cloud handles advanced reasoning and web searches.
• The system can mask sensitive information, keep it local, refuse an action or ask for user consent.
• Hybrid Compute protects highly confidential information such as credentials, payment card details and government-issued IDs.

Perplexity has introduced Hybrid Compute, a new feature designed to combine cloud-based AI with local processing on Apple Silicon-powered Macs.

The feature targets users who want advanced AI capabilities without sending sensitive information to the cloud. Instead, Hybrid Compute divides each task between a cloud model and a local AI model running on the Mac.

This approach gives users access to cloud-based reasoning and web search while keeping private information on their device when required.

How Perplexity Hybrid Compute Works

Perplexity uses an on-device privacy gate to decide what information can leave the Mac.

Before sensitive information reaches the cloud, the local Perplexity classifier checks the data. Depending on the situation, the system can mask sensitive details, keep the information on the device, refuse the action or ask the user for consent.

For example, highly confidential information such as credentials, payment card details and government-issued identification receives the strictest protection.

Meanwhile, the cloud can handle tasks that need more advanced reasoning, web searches and planning. The local model can handle private files, sensitive information and actions that need to remain on the Mac.

Perplexity Computer coordinates these processes automatically. Therefore, users do not need to manually decide which model should handle each part of a task.

Hybrid Compute Requires 24GB Memory

The feature currently targets relatively powerful Macs. Users need an Apple Silicon-powered Mac running macOS 15 or later, along with at least 24GB of unified memory.

Perplexity supports local models including Gemma 4 E4B, Qwen3.6 35B-A3B and a Perplexity model.

The company also says Apple’s Mac mini and Mac Studio are well suited to hybrid compute workloads.

However, the feature does not currently target every Mac user. Perplexity has made Hybrid Compute available to its Pro, Max and Enterprise subscribers.

How to Enable Hybrid Compute

Getting started requires the latest Perplexity Mac application.

Users can follow these steps:

1. Download and open the latest Perplexity app for Mac.
2. Download a supported local AI model with a single click.
3. Open the model selector and choose Hybrid.
4. Select the local and cloud models for the task.

Once configured, Perplexity Computer coordinates the workload between the selected local and cloud models.

A Different Approach to AI Privacy

Hybrid Compute reflects a growing push toward local AI processing. Running some AI workloads directly on a device can reduce the amount of sensitive information that needs to reach a cloud server.

At the same time, cloud models can still provide more advanced reasoning and access to online information. Perplexity’s approach attempts to combine both capabilities instead of forcing users to choose between local and cloud AI.

For professionals working with private documents and other sensitive information, this could make AI tools more practical for everyday workflows.

However, the feature currently comes with hardware and subscription requirements. Users need a compatible Apple Silicon Mac with sufficient unified memory, while access also requires a Pro, Max or Enterprise Perplexity plan.

Perplexity Hybrid Compute: Key Details