Perplexity announced a fully local mode for Perplexity Computer on supported Apple silicon Macs on October 20, 2026. The mode is for Pro, Max, and Enterprise subscribers, and Perplexity named Qwen 3.6 and PPLX Pro as local model options. It recommends 36 GB or more of unified memory for fully local use.
NeoTeo previously covered NVIDIA’s push to bring local AI into the home. Perplexity’s announcement adds a Mac-specific option built around its app.
What runs locally on the Mac
In fully local mode, the Mac runs the model, the agent directing the task, and the task execution. Perplexity says files used for that task are processed on the device rather than sent to cloud models. PPLX Pro is described by Perplexity as a post-trained version of Qwen.
With the user’s permission, Computer can work with local files and supported Mac apps. Perplexity names Notes, Messages, Mail, Numbers, Excel, and Keynote as examples. It also says code and tool execution take place in isolated sandbox environments with controlled access to files and connected apps.
Fully local mode and Hybrid Compute
Fully local mode and Hybrid Compute split processing differently. Perplexity recommends 36 GB of unified memory for fully local mode and 32 GB for Hybrid Compute. Both figures are recommendations, not stated minimums.
| Mode | Where processing happens | Recommended unified memory | Online work |
| Fully local | The Mac runs the model, agent, and task execution | 36 GB or more | Live web research and cloud features require an authorized cloud step. Sending messages and other online actions need a network connection. |
| Hybrid Compute | Cloud models handle research and reasoning; a local model processes private files | 32 GB | Cloud research and connected services require network access. |
For Hybrid Compute, Perplexity describes an on-device Privacy Gate that can keep personal information local, mask it before cloud use, refuse an action, or ask the user to approve sharing. Those controls apply to Hybrid Compute; fully local mode instead keeps the task’s model and processing on the Mac.
How the reported Mac configurations compare
AppleInsider reported these specific unified-memory configurations: the standard M6 Mac mini starts at 16 GB and tops out at 32 GB; M5 Pro Mac mini configurations include 48 GB and 64 GB; and the M5 Max Mac Studio starts at 36 GB.
| Mac configuration | Reported unified memory | Relation to the 36 GB recommendation |
| Standard M6 Mac mini | 16 GB to 32 GB | The reported maximum is below the recommendation. |
| M5 Pro Mac mini | 48 GB or 64 GB | Both reported configurations exceed the recommendation. |
| M5 Max Mac Studio | Starts at 36 GB | The reported starting configuration meets the recommendation. |
Perplexity’s 36 GB figure is guidance, not a hard compatibility cutoff. A configuration below it is below the recommendation, but that figure alone does not establish whether a particular Mac can run the mode.
Set up fully local mode
Perplexity’s setup sequence is:
- Install and open the latest Perplexity Mac app.
- Download and set up a local model in the app.
- Select fully local for the task.
When a task needs the internet or cloud
Local processing does not make network-dependent tasks work offline. Sending a message or meeting invitation still uses the relevant app’s service, and live web research requires an authorized cloud step.
Perplexity says users can authorize cloud escalation for current information, browser use, connected apps, or access to more than 15 frontier models. That option lets a task use cloud capabilities when needed; it is separate from processing the task entirely on the Mac.
