Perplexity Computer Now Runs Fully Local on Mac for Pro, Max and Enterprise. You Will Want 36 GB of Memory
The model, the agent and the code execution can all stay on your Apple silicon Mac. Hybrid mode is still there for web work and smaller machines. Here is what each mode needs and what it can reach.
TL;DR
Perplexity announced on 20 October 2026 that Perplexity Computer can now run fully locally on a Mac, with the model, the task planning and the code execution all staying on the device. It is available to Pro, Max and Enterprise subscribers in the Perplexity Mac app on Apple silicon, with 36 GB of unified memory or more recommended. Users pick one of two local models, Qwen 3.6 or PPLX Pro, a Qwen model Perplexity post-trained, and tasks can use Notes, Messages, Mail, Keynote and Numbers inside a sandbox. Hybrid Compute, launched on 1 September, remains for Macs with 24 GB or more and for jobs that need the web or one of more than 15 cloud models.
Perplexity launched Personal Computer for Mac in May 2026 and added Hybrid Compute on 1 September. The new step is a mode where nothing leaves the machine: the model that does the thinking, the agent that plans the steps and the sandbox that runs the code are all local.
What you need
- Plan: Perplexity Pro, Max or Enterprise.
- Hardware: an Apple silicon Mac. Perplexity recommends 36 GB of unified memory or more and points to the newest Mac mini configurations, including M5 Pro and M6 models, as good machines to leave running.
- Setup: update the Perplexity Mac app, follow the in-app steps to download a local model, then choose the fully local option when you start a task.
- Models: Qwen 3.6, or PPLX Pro, a version of Qwen that Perplexity trained further for agent work.
Fully local vs Hybrid Compute
| Fully local | Hybrid Compute | |
|---|---|---|
| Where the model runs | On your Mac | Frontier models in the cloud for research and reasoning, a local model for private files |
| Recommended memory | 36 GB or more | Works from 24 GB |
| Web browsing and live information | No | Yes |
| Cloud-connected tools | No | Yes |
| Local apps (Notes, Messages, Mail, Keynote, Numbers) | Yes, in a sandbox | Yes |
| Cloud models available | None unless you allow it | 15+ frontier models |
| Privacy layer | Data stays on device | Flags or masks personal details and asks before sending to the cloud |
The credit question
Perplexity frames the change as giving users more control over sensitive data and over credit spend. MacDailyNews goes further and says local runs do not consume Computer credits. Perplexity's own post is less direct about this, so check your usage page after a few local tasks before assuming they are free. Hybrid tasks that call cloud models still draw on credits.
What most coverage missed
- This is not a new product. Several top search results for Perplexity Computer still describe the May launch of Personal Computer, which needed the cloud for reasoning.
- Fully local means no web. If the task needs a search, a website or a cloud app, you are back in Hybrid mode.
- The 36 GB figure is a recommendation, not a hard limit, but a base 16 GB or 24 GB Mac is better served by Hybrid mode.
What this means for Mac users
- If you handle client files, health records or anything under a confidentiality agreement, fully local mode is the first Perplexity setup where the data path is simple to explain.
- If your Mac has less than 36 GB, stay on Hybrid and use the privacy layer's approval prompts.
- Expect local Qwen models to be slower and weaker than cloud frontier models on hard reasoning. Use local for private, repetitive file work and the cloud for open-ended research.
Sources
AI Industry Reporter
Priya covers model releases, industry announcements, and the gap between what labs claim and what independent evaluators actually find. She reads the primary source - the paper, the system card, the benchmark org's own statement - before writing a word.
More on AI Automation & Agents
Ai2's StateBench: Frontier Models Fall to 13% or Lower When Long Documents Keep Changing
Priya Nair · 4 min
Anthropic's Mythos 5 Report: The Model Kept Attacking a Real GitHub Project 98% of the Time Once It Started, Newer Claude Models Far Less
Priya Nair · 5 min
Anthropic Opens a Waitlist for Claude Duet, an Always-On Agent for US Pro and Max Subscribers
Priya Nair · 5 min