Perplexity Computer can now run fully local on Mac across private files and native apps

Perplexity is launching Computer fully local on Mac, allowing the AI model, harness, and task execution to run entirely on the device. With permission, the agent can work across local files and Mac apps such as Notes, Messages, Mail, Numbers, and Excel without sending personal data to the cloud, a move the company frames as giving users more control over sensitive data and credit spend. The feature is available now to Perplexity Pro, Max, and Enterprise subscribers in the Perplexity Mac app on supported Apple Silicon Macs, including the new Mac mini with M6 and M5 Pro. Perplexity recommends 36 GB of memory or more for local operation, and ships the mode with two local model options: Qwen 3.6 and PPLX Pro, described as a post-trained version of Qwen.

The launch gives Mac users a second way to run Computer alongside Hybrid Compute, which starts in the cloud and coordinates frontier reasoning with local processing of private files. Fully local mode inverts that: planning, reasoning, and file processing all stay on the machine, with users able to authorize escalation to the cloud for current information, browser use, and connected apps, at which point a local orchestrator can call on one of more than 15 frontier models. Perplexity illustrates the split with two contrasting scenarios: a portfolio analysis built on private holdings and downloaded market data can run entirely on-device, while a deal review combining live market research with confidential financials suits Hybrid Compute.

On security, Perplexity says code and tool execution run in isolated sandbox environments with controlled access to files and connected apps, and that the same protocols carry over to the cloud version of Computer. The company positions the release as the third step in a progression: bringing Computer to local files and native Mac apps with Personal Computer for Mac in May, then Hybrid Compute last month, and now full local execution. “As more demanding workflows become practical to run locally, the machines people already own will take on a larger share of their AI work”, Perplexity wrote.