Perplexity Hybrid Compute splits tasks between cloud AI models and on-device models running on Apple silicon Macs.
What the product does and how it is positioned
Perplexity Hybrid Compute enables the Perplexity Mac app to divide complex tasks across cloud and local compute resources. Cloud models conduct broad research and intensive reasoning, while an on-device model reads private files and runs sensitive steps locally on the user's Mac.
The system incorporates an on-device Privacy Gate classifier that detects sensitive information prior to outbound network calls. Personal data such as names, phone numbers, and addresses are substituted with placeholders for cloud processing and restored on the local machine upon completion.
Source-supported ways to use the product
Rebuilding internal financial projection models locally on a Mac while cloud models research comparable market transactions and return multiples.
Triggering analysis tasks remotely from an iPhone while routing confidential file reading and deck updates to an Apple silicon Mac.
The documented workflow, where available
Download and launch the latest Perplexity Mac application on an Apple silicon device.
Select and download a local model such as Gemma 4 E4B, Qwen3.6 35B-A3B, or a Perplexity model in one click without manual runtime configuration.
Open the model selector, choose Hybrid mode, and configure the desired local and cloud model pair for the task.
Hybrid Compute employs an open-source on-device PII classifier created with Perplexity's Secure Intelligence Institute. The classifier analyzes files and prompts locally on the Mac before external transmission takes place.
When personal data is detected, the Privacy Gate presents an option to process the document entirely on the local machine or upload it as is. If cloud processing proceeds, detected identifiers are swapped with stand-ins and later restored to the final output on the local device.
Checks to run with your own material and workflow
What was checked and when
Answers based on the source-checked product record
Hybrid Compute requires an Apple silicon Mac running macOS 15 or later with a minimum of 24 GB unified memory, while 32 GB is recommended for best results.
Users can download Gemma 4 E4B, Qwen3.6 35B-A3B, or a Perplexity model with a single click, requiring no Ollama installation or manual runtime setup.
An on-device PII classifier scans data before it leaves the Mac, replacing names, addresses, and account numbers with placeholders that are restored when cloud results return.
Yes, tasks can be initiated from connected devices such as an iPhone, after which the Mac runs the protected and local file-processing steps.
No, local model setup is handled in one click within the app without requiring an API key, Ollama, or manual runtime configuration.
