Perplexity Computer local mode is here: what it actually needs to run
On 20 October 2026, Perplexity launched the fully local mode for Perplexity Computer: the models and the agents run on your own Mac, not only in the cloud. The launch answers the 'what' - this page answers the question everyone asked next: does my Mac qualify?
What Perplexity actually announced
The announcement, made on 20-21 October 2026, introduced a local-first mode for Perplexity Computer on Apple Silicon Macs. Instead of every task being processed remotely, the models and the agent run on-device - the launch coverage named Qwen 3.6 and Perplexity's own PPLX Pro as the locally running models. Perplexity described the mode as offline-capable, which follows naturally: if the model and the agent live on your machine, a network connection is no longer the thing doing the work.
That shift is why hardware suddenly matters. In cloud mode, any Mac with a browser is equal. In local mode, your chip and your memory are the product - and Perplexity attached a specific hardware bar to the launch.
Where the 36GB requirement comes from
The announced requirement is at least 36GB of unified memory, on an Apple Silicon Mac (M1 or later). Both halves come straight from the announcement; neither is our invention. The full breakdown - including what has and has not been confirmed - is on our requirements page.
Why a memory floor at all? Running models and agents locally is memory-hungry: the model itself, its working context and the agent's in-progress tasks all live in unified memory, which on a Mac is one pool shared between the CPU and GPU. A 36GB floor is the smallest pool Perplexity is willing to support for that workload. Treat it as a minimum to clear, not a performance promise - a Mac at exactly 36GB qualifies, and a faster chip with more memory will still feel different in use.
Which Macs can run it - and which cannot
The rule produces some results that surprise people:
Can run it: any Apple Silicon Mac configured with 36GB or more - for example an M3 Pro or M3 Max at 36GB, an M1 Max or M2 Max at 64GB, an M4 Pro at 48GB, or any Ultra-tier machine. Age does not disqualify a Mac; memory does. We list every chip family and configuration in the full list of Macs that can run Perplexity Computer locally.
Cannot run it: every Intel Mac, regardless of RAM; every base M1, M2, M3 or M4 configuration (8GB-32GB); and - the painful one - 32GB machines, which miss the announced minimum by 4GB. If you are shopping, our Best Macs guide explains how to clear the 36GB line without overspending.
What has NOT been confirmed
We do not publish invented benchmarks, and the launch did not change that. There are no verified per-chip speed figures we can stand behind, no confirmed storage requirement on the evidence available, and no guarantee that every cloud feature behaves identically in local mode. Requirements can also change as the mode matures - always confirm the latest official requirements on Perplexity's own pages before buying hardware.
Frequently asked questions
When did Perplexity Computer local mode launch?
Perplexity announced the local-first mode on 20-21 October 2026. In this mode the models and agents run on your own Apple Silicon Mac instead of only in the cloud, and Perplexity described it as offline-capable.
What does my Mac need to run Perplexity Computer locally?
An Apple Silicon chip (M1 or later) and at least 36GB of unified memory. Intel Macs do not qualify, and neither do Apple Silicon configurations below 36GB - including 32GB.
My Mac has 32GB. Is that close enough?
No. The announced requirement is a 36GB minimum, and 32GB falls below it. Unified memory cannot be upgraded after purchase, so a 32GB Mac cannot be brought up to the line.
How do I check whether my Mac qualifies?
Run the free checker on the LocalAI Check homepage with your chip and memory, or see the full model-by-model list of which Macs can run Perplexity Computer locally.