Perplexity's local AI agent arrives on Windows with RTX 24GB VRAM requirement
Perplexity and Nvidia released Portable Computer for Windows, bringing the local AI agent to GeForce RTX and RTX PRO GPUs with 24GB or more of VRAM. The agent runs locally without consuming credits, but can optionally use cloud models with permission. New features include scheduled recurring tasks and local MCP servers for desktop apps.
The Windows release extends Portable Computer beyond its original Linux and DGX Spark environment, delivered through the existing Perplexity app. Qualifying hardware spans multiple generations, from the RTX 3090 through the RTX 5090, plus RTX PRO Blackwell workstation cards ranging from 24GB to 96GB of VRAM. The agent connects to productivity tools including Microsoft Word, Google Drive, Gmail, Slack, and GitHub, and supports one-click local model downloads such as Qwen 3.8 27B.
Nvidia also outlined broader plans at IFA, including RTX Spark Windows PCs from Lenovo and Acer arriving in October, alongside simplified local setup for other agents like Hermes Agent and OpenClaw. For existing RTX owners, the release removes the need to purchase a separate DGX Spark system, which carries a $4,699 price tag. A used qualifying GPU could offer a more affordable entry path for desktop users.
This release could lower the barrier to local AI agents for a niche but growing segment of PC owners with high-end GPUs, potentially shifting some AI workloads away from cloud services. Privacy-conscious professionals and researchers may benefit from keeping sensitive data on-device, while the credit-free local operation could reduce ongoing costs for heavy users. However, the 24GB VRAM requirement means most consumers remain excluded, possibly widening the gap between those who can afford capable hardware and those who cannot.