Perplexity Computer now runs fully on a Mac: a local PPLX Pro model post-trained from Qwen 3.8, no credits used, 36 GB of memory recommended

Why it matters

【Key points】

On Oct 20 (15:48 UTC), Perplexity said its agent product, Perplexity Computer, can now run fully locally on a Mac.

1. Both the model and the agent run on the user's Mac. It can work with native Mac apps and local files without sending task data to the cloud.

2. Local tasks use PPLX Pro, which Perplexity describes as its post-trained version of Qwen 3.8. Local tasks do not consume Computer credits.

3. A Hybrid mode can split a complex task between cloud models and the local model, keeping sensitive data on the device.

4. It is available to Pro, Max and Enterprise subscribers on supported Apple silicon Macs, with 36 GB of unified memory recommended, via the Perplexity Mac app.

【Verifiable】

Blog: https://www.perplexity.ai/hub/blog/perplexity-computer-now-runs-fully-local-on-mac

Launch post: https://x.com/perplexity_ai/status/2112571597798461920

Model details: https://x.com/perplexity_ai/status/2112571609693523990

【Our read】

1. Another agent product makes local-first a selling point: an open base model (Qwen 3.8), post-trained, in exchange for privacy and zero cloud cost.

2. The 36 GB recommendation puts it on high-end Macs only.

3. Perplexity has not published how the local model compares with its cloud models on task success, so complex-task quality is still unproven.