Project MonetRequest demo
Home/Blog/Perplexity Portable Computer: Local AI, DGX Spark, Models & Availability

AI · Project Monet Briefing

Perplexity Portable Computer Runs Its AI Agent Locally on NVIDIA DGX Spark

Perplexity’s Portable Computer moves its agent orchestration onto DGX Spark at launch, keeping many files and tasks local while allowing permissioned cloud escalation when needed.

Published 2026-08-29 · Updated 2026-08-29 · By Project Monet Editorial Team

Project Monet editorial diagram showing a local AI workstation processing private files with an optional permission-controlled path to cloud search, models and connectors

01

What Perplexity Portable Computer is

Perplexity announced Portable Computer on August 25, 2026 as a local-first version of its Computer agent. The launch puts the orchestrator, planner, tool router, scheduler, durable task queue and local search index on NVIDIA hardware instead of requiring every workflow step to run remotely.

Perplexity describes the system as hybrid rather than universally offline: local work can stay on the device, while tasks that need current web information, connected apps or stronger remote reasoning can escalate to cloud capabilities with user permission.

02

Launch hardware and operating system

The first release targets NVIDIA DGX Spark on Linux and is available to eligible Perplexity Pro and Max subscribers. DGX Spark is built around NVIDIA’s GB10 Grace Blackwell platform with a 20-core Arm CPU, NVIDIA GPU and 128 GB of unified memory.

03

Which local models does it use?

At launch, Portable Computer can use Qwen 3.8 27B or PPLX 27B. Perplexity describes PPLX 27B as a post-trained model designed for local knowledge work and for deciding when an authorized remote capability is useful.

Perplexity also announced NVIDIA Nemotron 3.5 Lightning as coming to the model picker. Treat that as announced future support until it is actually exposed in the released product.

04

What stays local

Perplexity says local models can read and search local files, work across documents and code, take actions on the device and keep jobs running. Dictation can also run locally using NVIDIA’s Nemotron 3.5 ASR model.

Code and tool execution are described as running inside isolated sandboxes with controlled access to files and connected apps. That is an important security boundary, but users should still review destructive actions and external writes.

05

When Portable Computer uses the cloud

A task can escalate to Perplexity search, research modes, browser tools, connected apps or more than 15 frontier models when the local system needs outside information or stronger remote reasoning. Perplexity says the user is asked for permission before device content is sent to a cloud service.

06

Connectors and external apps

Perplexity lists integrations including Google Drive, Gmail, Slack and GitHub. These connectors let a workflow combine local context with app data, subject to the permissions and authorization required by each service.

Use the minimum permissions needed for a workflow, especially when the local files contain confidential business material or private code.

07

Does local work consume Perplexity credits?

Perplexity says work handled locally does not consume credits. That can make repetitive or context-heavy workloads more practical on owned hardware, but it does not mean the product has no cost: launch access requires eligible Pro or Max membership and supported NVIDIA hardware.

Perplexity’s launch material does not publish a separate standalone Portable Computer price or a universal quota schedule, so a precise additional pricing table should not be fabricated.

08

How to get started on DGX Spark

Perplexity says eligible DGX Spark users can use a one-click installation flow through the Perplexity app. For the practical validation sequence, see how to install and use Portable Computer on DGX Spark.

09

Bottom line

Portable Computer is a hybrid local/cloud agent architecture: it keeps orchestration and many private-data tasks on DGX Spark at launch while allowing explicit escalation to current web information, connected applications and remote models. Windows, broader RTX support and Nemotron 3.5 Lightning in the picker remain announced future expansion rather than launch availability.

Sources

Primary and supporting sources

Facts were rechecked against the linked sources immediately before publication. Pricing, product availability and rollout status can change.

Project Monet

Useful signals. Clear decisions. Better digital work.

Project Monet turns relevant shifts in AI, creator tools and the web into practical context—and builds focused websites for businesses ready to grow.

Request a free homepage concept