Nvidia wants your home network to work like a mini data center for local AI
decoder ·
Nvidia has introduced PAIR, an open-source Personal AI Router that distributes local AI requests across available devices on a home network. It operates between tools such as Ollama or LM Studio and networked computers, allowing existing agents and applications to use multiple machines without modification. PAIR routes requests to available hardware, combines the results, and is designed to reduce delays for parallel agent tasks. It supports GeForce RTX 20-series and newer cards, RTX Pro workstations, DGX Spark systems, and Apple silicon devices beginning with the M4. In a demonstration, three devices completed a five-subagent task in just under nine minutes, compared with 18 minutes on one laptop. The tool automatically detects compatible devices, protects communication with mutual TLS encryption, and is available in beta for Windows, macOS, and Linux. Nvidia presents PAIR as part of its broader effort to connect open AI software more closely with its hardware ecosystem.
decoder