Technology
Nvidia releases open-source tool to distribute local AI tasks across network PCs
The Personal AI Router offloads parallel AI agent workloads to idle computers on a local network.
The short version
- Nvidia released the beta of Personal AI Router (PAIR), an open-source software tool that offloads AI subagent tasks to idle computers on a local network.
- The software targets users running local, multi-agent AI workflows on Windows, macOS, or Linux systems.
- PAIR routes individual parallel subagents to separate machines rather than pooling hardware memory or splitting a single model across multiple systems.
Key facts
- Nvidia announced PAIR, a free and open-source software tool released into beta on GitHub.[Engadget · The Verge · CNET]
- The tool functions by detecting local network devices via mDNS and assigning parallel subagent tasks to available machines to free up primary GPU capacity.[Engadget · CNET]
- Supported hardware includes Nvidia GeForce RTX 20-series or newer, Turing-generation or newer RTX PRO GPUs, DGX Spark systems, and Apple M4 chips or newer across Windows, macOS, and Linux.[Engadget · The Verge]
- PAIR currently integrates with the Ollama and LM Studio inference engines and uses MTLS for network security.[The Verge · CNET]
- The system assigns each subagent to an individual machine and does not combine or pool GPU memory across separate computers to run larger single models.[CNET]