Perplexity's Portable Computer runs AI locally for speed, privacy, and lower costs
Perplexity introduced Portable Computer, a local-first AI agent that processes tasks on-device without relying on cloud inference. This approach offers faster responses, enhanced security through an isolated sandbox, and eliminates cloud billing for local operations. Users can still connect to external services like Google Drive and Slack when needed, but the system requires powerful hardware for Linux or Windows.
The Portable Computer builds on Perplexity's earlier Personal Computer agent released in February, shifting the core processing from remote servers to the user's own device. This means files, models, and task execution remain on the machine, with the user deciding when the AI needs to go online. External services like Google Drive, Gmail, Slack, and GitHub can still be connected, but only the cloud-based portions of a task consume billing credits.
Availability is limited to paid subscription tiers—Pro, Max, Enterprise Pro, and Enterprise Max—with free users excluded. Linux systems get access first, while Windows support is promised for September. Hardware requirements are steep: Linux needs an Nvidia DGX Spark or an RTX GPU, and Windows demands an RTX GPU with at least 24GB of VRAM, which likely rules out many standard desktops.
This local-first approach could give privacy-conscious users and businesses greater control over sensitive data, reducing exposure to cloud breaches and eliminating inference fees. However, the hefty hardware requirements may restrict adoption to well-funded organizations or tech enthusiasts, potentially widening the gap between those who can afford privacy-enhancing AI and those who cannot. Smaller teams may find the upfront costs prohibitive, even as long-term savings accumulate.