Some AI agents that handle your email, notes, and private documents route that data through a remote server. Perplexity Computer now offers a different path, with the model, agent coordination, planning, reasoning, and task execution running directly on your Mac. If your Mac has at least 36GB of unified memory and you subscribe to Pro, Max, or Enterprise, this feature is available to you today.
Announced October 20, 2026, fully local Computer is available now to Pro, Max, and Enterprise subscribers through the Perplexity Mac app.
What Local Mode Actually Does
The entire agent pipeline runs on your Mac, not Perplexity’s servers.
When you trigger a local Computer task, the model and the system coordinating that task both operate on your device. Computer can read your Mail, scan your Messages, flag deadlines, and build an action list in Notes without that content leaving your machine, provided you grant the necessary Mac permissions.
Perplexity describes this as a meaningful privacy advantage for sensitive workflows. That claim comes from Perplexity directly and has not been verified through an independent security audit, so weigh it accordingly.
Data stays local unless you explicitly authorize cloud access. Tasks requiring live web information, browser access, or connected cloud services may require that authorization before content leaves the device.
Hardware, Models, and Who Gets Access
You need Apple silicon and at least 36GB of unified memory to use this feature comfortably.
Fully local Computer launches with two model options: Qwen 3.6 and PPLX Pro, which Perplexity describes as its own post-trained version of Qwen. Both run through the Perplexity Mac app rather than requiring a separate local inference setup.
Perplexity recommends 36GB of unified memory or more, placing the feature in MacBook Pro and Mac Studio territory for most users. Macs with less memory are not listed among the supported configurations.
Setup is straightforward: download and open the latest Mac app, download a local model from within the app, and select the fully local option when starting a Computer task. Local tasks do not consume Computer credits, which matters if you use Computer regularly.
Availability is limited to Pro, Max, and Enterprise subscribers. Perplexity does not list free-tier access for this feature.
The Difference Between Local and Hybrid
Hybrid Compute blends cloud reasoning with local file handling, while fully local keeps everything on your Mac.
Perplexity also offers Hybrid Compute, and the distinction is worth understanding before you choose a mode. Hybrid Compute splits a task between cloud models and a local model. Cloud handles complex research and reasoning; the local model processes your private files and Mac applications.
Think of Hybrid Compute as the option for tasks where you need current information but want sensitive file processing kept close. Fully local suits situations where connectivity is unreliable or the task is self-contained, though users can still authorize cloud access when needed for specific steps.
What This Means in Practice
Local performance depends on your specific Apple silicon Mac, and cloud-dependent tasks still require an explicit authorization.
Your results will depend on the Mac in front of you. Local model performance scales with Apple silicon capability and available unified memory.
Whether local models can consistently match cloud systems on reliability and reasoning quality remains an open question. Perplexity’s positioning suggests that for routine personal and professional workflows, supported Macs are already capable enough to handle the job, and this launch represents a clear step toward putting more of that work directly on your device.
