Sensitive local knowledge work
Teams analyzing private files or code that should remain on a controlled machine for the local portion of the task.
Independent tool overview
Portable Computer at a glance
Perplexity Portable Computer is a local-first version of Perplexity Computer that runs its agent, models, search index and tool orchestration on supported NVIDIA hardware, with permission-based cloud escalation when a task needs outside data or stronger models.
Visit the official Portable Computer site ↗
- Best for
- private, high-volume agent work on supported local NVIDIA hardware
- Launch platform
- NVIDIA DGX Spark running Linux
- Local models
- Qwen 3.8 27B and Perplexity's PPLX 27B
- Access
- Perplexity Pro or Max subscription plus compatible hardware
Overview
What Portable Computer is
Portable Computer launched on August 25, 2026 as a way to run Perplexity's agentic workflow locally for private files, codebases and high-volume tasks. Local inference, planning, tool routing, scheduling, task queues and document search run on the user's machine instead of consuming cloud credits.
The first release is for NVIDIA DGX Spark on Linux and requires a Perplexity Pro or Max subscription. Perplexity says support for NVIDIA RTX GPU PCs and Windows is coming, so most existing laptops and desktops are not yet eligible for the launch version.
Portable Computer is local-first, not necessarily local-only. It can ask permission to use Perplexity search, research, browser capabilities, connected apps or frontier cloud models. Users must review that boundary carefully because an approved escalation can move selected task content off the device.
Use cases
Who Portable Computer is best for
The strongest fit depends on the job you need the product to complete, not the size of its feature list.
High-volume recurring tasks
Users who own capable hardware and want repeated local inference without a per-credit charge for on-device work.
Hybrid local and cloud research
Workflows that keep confidential source material local but occasionally need current web data or frontier reasoning with explicit approval.
Capabilities
Core Portable Computer features
Local agent stack
Runs the orchestrator, planner, tool router, scheduler, durable task queue and local search index on the device.
Local file and code actions
Reads, searches and acts on files and code using on-device models and sandboxed tool execution.
Permission-based cloud escalation
Requests user approval before sending relevant content to cloud search, browser tools, connected apps or advanced models.
Local dictation
Uses NVIDIA's Nemotron ASR model for on-device speech transcription so audio can stay on the machine.
App connectors
Can work with services such as Google Drive, Gmail, Slack and GitHub when the user authorizes the connection and data movement.
Process
How the Portable Computer workflow works
Step 1
Classify the data boundary
Decide which files, credentials and actions must stay local and which tasks may use approved cloud services.
Step 2
Run locally first
Give the agent a bounded task, inspect requested file access and let the local models complete as much work as possible.
Step 3
Review every escalation and action
Check what content will leave the device, which service will receive it and what action the agent plans before granting permission.
Cost
Portable Computer pricing and free plan
Portable Computer does not have a separate published software price, but it requires a Perplexity Pro or Max subscription and supported NVIDIA hardware. Local tasks do not consume Perplexity Computer credits. Cloud escalations can use credits, and the NVIDIA DGX Spark launch hardware is currently listed at $4,699 in the U.S.
Perplexity Pro
$20/month or $200/year
Lowest subscription tier currently eligible for Portable Computer.
- Requires compatible NVIDIA hardware
- On-device work does not consume credits
- Consumer Pro has no recurring monthly Computer credit allocation
- Cloud Computer usage may require purchased credits after any promotional balance
Perplexity Max
$200/month or $2,000/year
Power-user subscription with a recurring Computer credit balance and higher Perplexity limits.
- 10,000 Computer credits per month
- Unused monthly credits do not roll over
- On-device work does not consume credits
- Cloud escalation and other Computer tasks can consume credits
NVIDIA DGX Spark
$4,699 one-time hardware purchase
The supported desktop system for the first Portable Computer release.
- 128 GB unified system memory
- 4 TB NVMe storage
- NVIDIA GB10 Grace Blackwell platform
- Linux launch support; Windows support is announced as coming
Pricing checked . Check current pricing at the source ↗
Assessment
Portable Computer strengths and limitations
Where it stands out
- Keeps the core agent loop, local index and supported file work on hardware controlled by the user.
- Local tasks avoid the variable cloud-credit charge associated with Perplexity Computer.
- Hybrid escalation preserves access to current web information and stronger cloud models when explicitly approved.
What to consider
- The launch version requires an expensive DGX Spark and Linux, sharply limiting who can use it immediately.
- RTX PC and Windows support are announced but not yet available in the initial release.
- Local 27B models may be less capable than frontier cloud models on demanding reasoning, coding or research tasks.
- Local-first does not mean no data ever leaves the device; cloud escalations and connectors can transmit approved content.
- Agent access to files, apps and credentials creates operational risk even when inference stays local.
Compare
Portable Computer alternatives
The right alternative depends on the specific output, workflow, controls and budget your project requires.
Agents
Perplexity Computer
Choose Perplexity Computer when cloud execution and broad device access matter more than keeping the agent stack on local NVIDIA hardware.
Explore Perplexity Computer →Agents
Perplexity Personal Computer
Choose Perplexity Personal Computer for local-file and native-app workflows on a normal supported PC without the launch version's fully local model stack.
Explore Perplexity Personal Computer →Agents
OpenClaw
Choose OpenClaw when an open, self-managed agent architecture matters more than Perplexity's packaged local-and-cloud workflow.
Explore OpenClaw →Questions
Portable Computer FAQs
What is Perplexity Portable Computer?
It is a local-first version of Perplexity Computer that runs models, planning, tool routing, task state and local search on supported NVIDIA hardware.
What hardware does Portable Computer require?
The first release supports NVIDIA DGX Spark. Perplexity and NVIDIA say support for RTX GPU PCs is coming, but it is not part of the initial launch availability.
How much does Portable Computer cost?
There is no separate published software fee. You need Perplexity Pro at $20 monthly or Max at $200 monthly plus compatible hardware; NVIDIA currently lists DGX Spark at $4,699 in the U.S.
Does Portable Computer work offline?
Its local models and core agent stack can handle on-device work, but web research, connected apps and frontier-model escalation require network access and user authorization.
Does all data stay local?
Local tasks keep their data on the machine. When a task needs a cloud service, Portable Computer asks permission before sending the relevant content off-device, so the user must review each escalation.
Bottom line
Our Portable Computer verdict
Portable Computer is an important early example of a polished local-first agent, but today it is a specialist product for people who already want a DGX Spark-class machine. The economics and privacy story become compelling only when a meaningful volume of work can stay local and the organization has strong controls around file access, credentials and cloud escalation.
Visit Portable Computer website ↗Similar Consumer Tools
View all Consumer AI tools →
ChatGPT for Teens
ChatGPT for Teens - OAI's new teen experience pairing guided learning with default safety limits
GLM-5.3-Flash
Z AI’s new multimodal system revealed as the mystery Ox Alpha model on OpenRouter
GLM-5.3
GLM 5.3 - Z AI’s new open-source model with strong coding, agentic, and cyber capabilities
Gemini 3.8 Flash
Google's strongest Flash model yet for coding and agents
AI training for the future of work.
Get access to all our AI courses, hundreds of real-world AI use cases, live expert-led workshops, an exclusive network of AI early adopters, and more.
AI Courses
Get unlimited access to all of our current & upcoming industry-specific AI courses for the duration of your subscription.
Daily Guides
To keep up with the rapid pace of AI, our team publishes AI implementation guides daily. Our library contains 300+ practical use cases to automate real-world work.
Workshops
Join weekly, live, interactive sessions with industry leaders who are at the forefront of AI for hands-on implementation guidance and exclusive insights.
Community
Network with an exclusive community of AI-first professionals who are working smarter with AI. Learn how early adopters are using AI in their work and businesses.