Perplexity launches privacy-focused hybrid compute for Apple silicon

Perplexity is launching its new hybrid compute feature for Mac, allowing AI tasks to be split between cloud models and a local model running on Apple silicon. The system begins processing in the cloud, but sensitive steps are handled locally on the user’s Mac to improve privacy. Before anything is sent online, the Mac scans the text and hides personal details like names, addresses, and account numbers by replacing them with temporary placeholders. Perplexity has open-sourced this privacy tool, developed with its Secure Intelligence Institute. Hybrid compute requires at least 24GB of unified memory, with 32GB recommended, and works on macOS 15 or later. Users can download the PPLX Qwen 3.8 27B model with one click, without needing external tools. Tasks handled locally do not consume cloud credits or require API keys. The feature builds on Perplexity’s recent Mac app overhaul and is highlighted by Apple as a productivity use case for the new M6 Mac mini.

Read the full story on 9to5Mac →