Perplexity brings local AI processing to Apple Silicon Macs

To achieve seamless performance, Perplexity recommends at least 36GB of unified memory. The local system utilizes either Qwen 3.6 or PPLX Pro—a specialized post-trained iteration of the model—to handle complex queries directly on the device. By keeping data local, the application offers a secure environment for processing private documents and internal application tasks that users may prefer to keep off external servers.

Despite the offline capability, the system remains hybrid by design. Users retain the flexibility to shift specific tasks to the cloud when they require real-time web access or updated information. This tiered approach allows the software to balance the high-security demands of local processing with the broad, current knowledge base provided by Perplexity’s standard online infrastructure.