Perplexity AI workflows have reached a major milestone: you can now run the entire Computer experience locally on your own Mac. This means the harness, the orchestrator and everything in between can stay on your device instead of being routed through the cloud. The company introduced Hybrid Compute in early September 2026, and this full local option is the natural next step.
The catch is hardware. You’ll need a recent Apple Silicon Mac with at least 36GB of RAM to take advantage of it. If your machine clears that bar, Perplexity AI workflows can run start to finish without leaving your computer.
Why Run Everything Locally?
Perplexity is pitching local execution as a win for two groups of users: those watching their credit spend and those who want to keep sensitive data private. The company’s own example makes the appeal clear. With access to your Notes and Mail apps, you could ask Computer to generate a priority to-do list for you—without any of the data in your emails being sent to the cloud.
“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 explains. “This gives users more control over sensitive data and credit spend.”
That framing matters. For many users, the biggest barrier to heavy AI use isn’t capability—it’s cost and trust. Running Perplexity AI workflows on local hardware addresses both at once.
Two Models to Choose From
When you run Computer locally, you choose between two models:
Qwen 3.6 — the standard local option
PPLX Pro — a Perplexity offshoot of Qwen 3.6 that the company post-trained specifically to fine-tune it for Computer use
The second option is notable because it shows Perplexity isn’t just shipping an off-the-shelf model. The company did its own post-training to adapt Qwen 3.6 for the way Computer actually works.
You Can Still Call the Cloud When Needed
Local doesn’t mean isolated. If you feel a task would benefit from a frontier system chipping in, you can authorize the software to call on one of more than 15 models. That keeps the hybrid approach available even when you start with a local model, so you’re not locked out of extra capability when a job demands it.
Security and Sandboxing
Any code or tool execution a task requires is done in an isolated sandbox environment. Perplexity notes that “these same security protocols persist in the cloud version of Perplexity Computer,” which means the safety model doesn’t weaken just because the compute moved onto your machine.
For privacy-focused users, this is the core selling point. Sensitive data stays local, and the execution environment remains contained.
Who Gets Access and How to Set It Up
Perplexity Pro, Max and Enterprise subscribers can begin using Computer locally starting today. Setup is straightforward:
Download and install the latest version of the Perplexity Mac app.
Follow the in-app instructions to download and install a local model on your machine.
Choose between Qwen 3.6 and PPLX Pro, then start running tasks.
There’s no complicated configuration process to work through—the app walks you through the model download directly.
What About the RAM Requirement?
The 36GB RAM floor is the biggest practical hurdle. Many recent Apple Silicon Macs ship with 8GB, 16GB or 24GB, so a large portion of the user base won’t qualify yet. If you’re considering upgrading hardware specifically for local AI work, this requirement is worth factoring in.
Perplexity AI workflows running entirely on-device signal where the company sees the future heading. Local execution is moving from novelty to practical option, and the pitch is consistent: more control over data, more control over cost, and the hardware you already own doing more of the work.
Perplexity’s own framing captures it well: as more demanding workflows become practical to run locally, the machines people already own will take on a larger share of their AI work. For subscribers with capable Macs, the entire Computer experience no longer has to leave the device.
The trade-off is clear. You get privacy and potentially lower credit spend, but you need serious hardware to participate. Whether that balance works for you depends on what you’re running and how much you value keeping it local.
