Perplexity’s hybrid compute on Mac enhances privacy and saves costs with split-processing AI model

Perplexity has launched Hybrid Compute for Mac, a new feature that splits AI workloads between local models and cloud-based systems, enhancing data privacy while reducing costs for users with Apple silicon Macs.

Perplexity has added a new Mac feature called Hybrid Compute, extending its Computer product with a split-processing model that sends part of a task to cloud-based AI and part to a model running locally on the user’s Mac. The company says the aim is to keep sensitive material on the device while still using more powerful cloud systems for other steps, a design that also lets users shift some inference work away from higher-cost frontier models.

According to Perplexity and coverage from 9to5Mac, the feature is intended for workflows that involve private files or other confidential data. Before a task is broken up, the user can review which files the system plans to keep local. Perplexity says the Mac app handles local-model installation automatically, so there is no need to use Terminal or carry out manual setup steps.

The company is also pitching Hybrid Compute as a cost-control measure. By offloading some work to a local large language model, users can reduce the amount of cloud inference they consume, while the app shows device resource use and token consumption as the task runs. Perplexity also says follow-up prompts are supported after a response is generated, and tasks can be queued from an iPhone.

Perplexity has limited the feature to Macs with Apple silicon running macOS 15 or later, and it recommends at least 24GB of unified memory, with 32GB described as preferable. The feature is available to Pro, Max and Enterprise subscribers. Reporting from MacRumors and other outlets also places the launch in the wider context of Perplexity’s earlier Mac products, including Personal Computer, which broadened Computer’s access to local files and native apps.