Perplexity Computer users on Mac can now split every job between a frontier model in the cloud and a model that runs locally on their machine.
The mode launches for Pro, Max, and Enterprise subscribers with three local options: Gemma 4 E4B, Qwen3.6 35B-A3B, and a Perplexity model post-trained for Computer. It requires any Apple silicon Mac with macOS 15 or newer and at least 24GB of unified memory.
The cloud handles reasoning, search, and planning while the local model processes private files and sensitive actions. Nothing leaves the machine without passing a privacy gate first. The gate can mask names, addresses, account numbers, and secrets, or refuse a request outright. Credentials, card numbers, and government IDs get the strictest treatment, with values swapped for stand-ins and restored when answers return.
Perplexity open-sourced the classifier behind the gate, PII-Tracer, a 0.6B-parameter model built on a Qwen3 backbone, along with a benchmark of 13,148 conversations across 13 languages.
Work handled by the local model uses no cloud credits, and enterprise admins can set organization-wide rules for what stays on the Mac. The company points to Apple’s new Mac mini and Mac Studio as ideal hosts, with an always-on Mac mini reachable from an iPhone.
