Portable Computer is a local build of Perplexity's agentic Computer platform that runs entirely on a user's own machine, starting with NVIDIA's DGX Spark desktop system and expanding to RTX GPU PCs. The orchestrator LLM, subagent models, and agent harness all execute on device, so analyzing files, running research, and driving connected apps like Gmail, Outlook, Slack, and GitHub happen without a round trip to the cloud. This launch matters because it's aimed at Pro and Max subscribers who already have serious hardware and want to run agents on private data without paying per token for every step.
Under the hood, Portable Computer runs a post-trained PPLX 27B model, with Qwen 3.8 27B also available and NVIDIA Nemotron 3.5 Lightning support coming soon. When a task needs frontier reasoning, the orchestrator asks for approval before routing that single step to a cloud model, gated by the user, flagged for PII, and limited to text guidance so sensitive documents stay on the device. Code and tool execution run in isolated sandbox environments with controlled access to files and connected apps , matching the security posture of the cloud product.
The initial hardware bar is real. The easiest path is a DGX Spark desktop running DGX OS, but a traditional PC on Ubuntu with an NVIDIA RTX card carrying at least 24 GB of VRAM also works. A DGX Spark costs around $4,800, and even an older RTX 3090 with 24 GB of VRAM runs well over $1,500. The first release is Linux only, with Windows support coming soon, and installation is a one-click setup from the Perplexity app. For teams already sitting on NVIDIA workstations, it's a concrete way to move repo-scale migrations, document synthesis, and long-running agent workflows off metered cloud infrastructure.
Comments (7)
Sign in to join the discussion.
N
Nia AdesolaAug 27, 2026
Wait, is it Portable Computer or Personal Computer? The tagline and the tweet are having two different launches right now.
T
Tomas KowalskiAug 27, 2026
Fully local until you check the network tab and see the phone-home for 'telemetry'. Prove me wrong on the DGX.
S
Shan R.Aug 27, 2026
That cold open with the terminal boot sequence is elite editing, but the second act drags for like 15 seconds before the agent actually does anything. Trim it.
V
Vikas PatnaikAug 27, 2026
Cool, so the total addressable market is people who own a DGX Spark. Truly launching for the dozens.
K
Kenji O.Aug 27, 2026
Is there a local endpoint I can hit from my own harness, or is the orchestrator locked to your UI? Asking for every dev who wants to swap the subagent LLM.
M
Mariel HaddadAug 27, 2026
General counsel brain activated: 'no cloud dependency' is the phrase every regulated buyer has been waiting to screenshot into a procurement deck.
L
Lars BergqvistAug 27, 2026
Been prototyping something adjacent for 3 weeks and this just quietly ate half my roadmap. Back to the whiteboard I guess.
Work With Us
Book an intro call to explore our suite of services. We might have a waitlist.
Perplexity
Personal Computer is here
Personal Computer is here
Tags
TLVC Rating
Community Rating
No ratings yet
to rate this launch.
Portable Computer is a local build of Perplexity's agentic Computer platform that runs entirely on a user's own machine, starting with NVIDIA's DGX Spark desktop system and expanding to RTX GPU PCs. The orchestrator LLM, subagent models, and agent harness all execute on device, so analyzing files, running research, and driving connected apps like Gmail, Outlook, Slack, and GitHub happen without a round trip to the cloud. This launch matters because it's aimed at Pro and Max subscribers who already have serious hardware and want to run agents on private data without paying per token for every step. Under the hood, Portable Computer runs a post-trained PPLX 27B model, with Qwen 3.8 27B also available and NVIDIA Nemotron 3.5 Lightning support coming soon. When a task needs frontier reasoning, the orchestrator asks for approval before routing that single step to a cloud model, gated by the user, flagged for PII, and limited to text guidance so sensitive documents stay on the device. Code and tool execution run in isolated sandbox environments with controlled access to files and connected apps , matching the security posture of the cloud product. The initial hardware bar is real. The easiest path is a DGX Spark desktop running DGX OS, but a traditional PC on Ubuntu with an NVIDIA RTX card carrying at least 24 GB of VRAM also works. A DGX Spark costs around $4,800, and even an older RTX 3090 with 24 GB of VRAM runs well over $1,500. The first release is Linux only, with Windows support coming soon, and installation is a one-click setup from the Perplexity app. For teams already sitting on NVIDIA workstations, it's a concrete way to move repo-scale migrations, document synthesis, and long-running agent workflows off metered cloud infrastructure.
Comments (7)
Wait, is it Portable Computer or Personal Computer? The tagline and the tweet are having two different launches right now.
Fully local until you check the network tab and see the phone-home for 'telemetry'. Prove me wrong on the DGX.
That cold open with the terminal boot sequence is elite editing, but the second act drags for like 15 seconds before the agent actually does anything. Trim it.
Cool, so the total addressable market is people who own a DGX Spark. Truly launching for the dozens.
Is there a local endpoint I can hit from my own harness, or is the orchestrator locked to your UI? Asking for every dev who wants to swap the subagent LLM.
General counsel brain activated: 'no cloud dependency' is the phrase every regulated buyer has been waiting to screenshot into a procurement deck.
Been prototyping something adjacent for 3 weeks and this just quietly ate half my roadmap. Back to the whiteboard I guess.