Perplexity Brings Its Local Portable Computer Agent to Windows PCs

Perplexity made Portable Computer, the local on-device version of its Perplexity Computer agent, available in the Perplexity app for Windows on September 14, 2026, on compatible NVIDIA GeForce RTX PCs and NVIDIA RTX PRO Workstations with 24GB or more of VRAM.

Availability for Windows Subscribers

Portable Computer for Windows is available to Pro and Max subscribers on both individual and enterprise plans through the existing Perplexity app for Windows, which is available for download through the Microsoft Store, according to Perplexity’s announcement. After downloading the app, users download the local model they want to use in one click after selecting it in the model dropdown.

On-device inference requires an NVIDIA GeForce RTX or RTX PRO GPU with at least 24GB of VRAM. NVIDIA said support for its DGX Station workstation is expected to come soon.

From DGX Spark to the Windows Desktop

Portable Computer is a local version of Perplexity Computer, an agent that plans and carries out multistep tasks. Accelerated by NVIDIA GPUs, it uses local models to analyze data, bring together information across files, and handle recurring work, while sensitive information stays on the device and locally completed work does not consume Perplexity Computer credits, according to NVIDIA’s announcement.

Perplexity introduced Portable Computer in August for the NVIDIA DGX Spark, a desktop-sized AI computer built around the Grace Blackwell GB10 platform with a 20-core Arm CPU, an NVIDIA GPU, and 128GB of unified memory. Perplexity built the release with NVIDIA and launched it first on Linux, with Windows support promised at the time.

Before the DGX Spark release, Perplexity’s Personal Computer for Windows had brought Computer’s cloud agents to the Windows desktop, where they could work across local files, connected apps, and the web. Portable Computer instead runs the model, agent harness, orchestrator, and scheduler entirely on the Windows device. NVIDIA said the Windows release builds on existing support for DGX Spark systems and RTX PCs running Linux.

Local Models, Connectors, and Scheduled Tasks

NVIDIA said the Windows app simplifies setup with a local model such as Qwen 3.8 27B, post-trained to work with Perplexity Computer and optimized for RTX GPUs, so users can put the agent to work without researching models or configuring the complex software stack typically required to run local AI.

On the DGX Spark release, Portable Computer runs Qwen 3.8 27B or PPLX 27B, a post-trained version of the Qwen model, while the orchestrator, planner, tool router, scheduler, durable task queue, and local search index all run on device. Perplexity said NVIDIA Nemotron 3.5 Lightning, a 30B open model, is coming soon to the model picker, and that dictation mode runs locally with the NVIDIA Nemotron 3.5 ASR model, keeping transcription and file actions on the machine.

Connectors for Microsoft Outlook, OneDrive, Word, Google Drive, Gmail, Slack, and GitHub extend the agent across the files and apps already part of users’ daily workflows, according to NVIDIA. Portable also connects to other desktop applications through local Model Context Protocol servers running on the Windows device, which Perplexity said lets the local model use those applications’ tools and data as part of an automated workflow while keeping the work on device. The agent can work with documents, spreadsheets, presentations, PDFs, code, and images stored in permitted folders, then use tools to act on that information.

Users can schedule recurring workflows, such as analyses or file processing, to run with a local model on their own hardware. Perplexity described a controller at a logistics company scheduling Portable to reconcile freight invoices in a permitted folder each morning: the agent matches each invoice against the carrier’s local rate sheets, flags duplicate charges or rates that deviate from contract, and generates an exception queue showing the invoice, the discrepancy, and the recommended dispute action, with the invoice and rate data never leaving the company’s device.

Cloud Escalation and Security

For tasks that need current information or advanced reasoning, Portable Computer can include results from Perplexity Search or one of 15+ frontier models, and it asks the user’s permission before sending information off the device. Perplexity said the agent can also call Perplexity Search and run wide or deep research, then return the findings to the same task.

Perplexity Computer’s built-in tools include a built-in browser and the company’s proprietary SPACE sandbox, according to NVIDIA. Perplexity said code and tool execution run in isolated sandbox environments with controlled access to files and connected apps, and that security matches the cloud version of Perplexity Computer.

NVIDIA’s post outlined example workflows for the agent. An engineer can review open pull requests in a connected GitHub project, organize them by status, and identify next steps, with the agent also able to flag outdated documentation and submit proposed updates for review. A finance user can point it at two years of brokerage summaries, consolidated 1099s, and tax returns and have it trace the recurring holdings creating the most avoidable fees and tax drag, with every figure cited to the exact file and page and no document reaching a chatbot. A startup can ask why activation went flat and have the agent analyze the funnel export locally to find where new signups drop off between install and first completed task, then post the top insights to the team’s Slack channel.

Perplexity gave its own engineering example: a user can ask Portable to review open pull requests against the local codebase and group them as ready to merge, needs review, blocked, or stale, and after the user approves the summary and action items, the agent posts them to the team’s Slack channel and tags the directly responsible individual for each item.