Video wird geladen...
Video konnte nicht geladen werden
Perplexity Computer can now run fully local on your Mac. Work with your native Mac apps and local files, and run tasks without sending sensitive data to the cloud. The model and the agent both run on your Mac. Available now in the Perplexity Mac app.
326,351 Aufrufe • vor 1 Tag •via X (Twitter)
30 Kommentare

Local tasks use PPLX Pro, our post-trained version of Qwen 3.8, and don't consume Computer credits. For more complex work, Hybrid can split the task between a cloud models and a local model on your Mac, keeping sensitive data private. Read more:

Available now to Pro, Max, and Enterprise subscribers on supported Apple silicon Macs. 36 GB of unified memory is recommended. Download the Perplexity Mac app and set up a local model to get started.

Love that we can work with local files now without worrying about cloud data privacy.

The Mac becoming the whole AI workspace is the part that lands. Sensitive files stay on the machine, and the assistant still gets useful work done. Where does it feel fastest first, native apps or local files? Best, Pete

Perplexity is changing the open source game day after day

Local is the right call for the boring stuff people actually want automated: spreadsheets, mail, files with client data. The real question is how big the on-device model is, and whether it quietly falls back to the cloud when a task gets hard, or asks first.

Whoa, nice for anyone with sensitive files on their laptop. any chance it comes to linux too? 🐧

running both the agent and model local without telemetry is huge for enterprise folks who were terrified of data leaks

rip to every chrome tab i have open

Local agents on Mac finally make sense for sensitive files

does it still work with wifi off

a local agent on a mac is the first ai pitch that gets the it department to stop sweating

Inbox full of dates you keep meaning to write down? With permission, you can ask Computer to pull deadlines from your recent Mail into a Notes to-do list, fully on your Mac. For Pro, Max and Enterprise on Apple silicon Macs; 36 GB+ memory recommended.

Amazing work, this makes local workflow integration so much easier.

Does local mean the agent can read local files without touching iCloud?

i think this is a big unlock all the service providers are converging into this primitive having an always-on orchestrator that can access your personal computer on demand

is the local agent running inside the SPACE sandbox you open sourced last week? if yes thats the real story, anyone can read exactly what it's allowed to touch on my machine

Awesome!!!

Model and agent both on the Mac. The fan just got a new job.

hybrid is the part I would want spelled out, who decides which piece of a task counts as sensitive before it leaves the mac

an ai agent that runs fully on your mac, touches your local files, and doesn’t ship sensitive data to the cloud? that’s the version people have been asking for 📁

finally an agent I'd let into my downloads folder, mostly because nobody else will ever see what's in there

The 36GB recommendation is the real filter here, most 16/24GB MacBooks are out. Curious how the cloud handoff works: a prompt every time, or preset rules per task type?

What? In 2026? We already use codex, Gemini app, Hermes...etc..so many options

36GB recommended means my 16GB Mac would keep my data private mostly by never finishing the task.

Both the model and the agent staying on the Mac is the part that matters for sensitive files. Which Macs run it comfortably?

Local agents on Mac finally make sense for sensitive files

Amazing!

fully local is what gets agents past the security review. most companies say no to cloud agents long before anyone asks how smart the model is

Fully local is the only kind of AI I'd trust with my files

