Perplexity unveils Portable Computer, a local AI agent for faster, private tasks
Perplexity launched Portable Computer, a local-first AI agent that runs models on-device, offering speed, security and cost savings. It's available now for paid Linux users with Nvidia RTX GPUs, coming to Windows in September.
The agent, built on the company's February-released personal computer tool, handles standard AI commands and agentic tasks with little or no user intervention. All models, files and work stay on the user's machine, and the AI only goes online when the user permits, such as when connecting to external services like Google Drive, Gmail, Slack or GitHub.
Perplexity said in a blog post that sensitive data never leaves the device without permission, and local models carry no inference fee. "The system is private and cost-effective by construction," the company said.
The new option is available only to paid subscribers, including Pro, Max, Enterprise Pro and Enterprise Max plans. Free users are not eligible. Initially, Portable Computer runs only on Linux systems, with Windows support promised for September.
Running local AI requires serious hardware. On Linux, users need an Nvidia DGX Spark or another Linux machine with an Nvidia RTX GPU, running Nvidia DGX OS or Ubuntu on ARM or x64 systems. On Windows, the PC must have an Nvidia RTX GPU with at least 24GB of VRAM. Such graphics cards typically cost at least $1,500, which could exclude many desktop users.
For local models, users can choose Qwen 3.8 27B, a well-known open-source model praised for speed, coding, research and complex agentic tasks, or a post-trained version called PPLX 27B. Perplexity post-trains all models on Portable Computer to boost accuracy and efficiency. Nvidia's Nemotron 3.5 Lightning, designed for high-volume and long-running tasks, is coming soon, and users will be able to switch among models depending on their assignments.
Whether a user's system can handle the heavy demands or not, Portable Computer represents a new option for local AI, and the hardware requirements may ease over time, potentially making local AI more accessible for better performance, security and cost savings.