NVIDIA Releases PAIR, a Free Tool That Routes AI Tasks Across Your Local Network
PAIR turns idle machines on a home or office network into a pooled inference cluster, addressing the GPU bottleneck that multi-agent workflows create on a single machine.
Reporting from 1 source: GIGAZINE.
NVIDIA has released PAIR (Personal AI Router), a free virtual inference router that detects compatible PCs on a local network and assigns AI inference jobs to them. It works with Ollama and LM Studio and needs no new API. In a demo running five subagents with Qwen 3.6-35B-A3B, a three-machine cluster averaged 8 minutes 48 seconds versus 18 minutes on a single RTX Spark PC.
PAIR's proxy is compatible with Ollama and LM Studio, so it requires no new API. When no jobs are queued, it powers off or hibernates idle nodes.
The demo paired PAIR with Hermes Desktop, a GUI for the Hermes Agent, and Ollama to route five subagents. With Alibaba's Qwen 3.6-35B-A3B, a cluster of an RTX Spark PC, a DGX Spark workstation, and an RTX 5090 PC averaged 8 minutes 48 seconds, against 18 minutes on a single RTX Spark PC.
Requirements cover Windows 11, DGX OS, Ubuntu 14.04, or macOS Tahoe, a GeForce RTX 20 series or newer, DGX Spark, or Mac M4 or newer, 8GB of memory, and 20GB of free space.
Synthesized by Yomimono from the 1 cited source below, including Japanese-language reporting where cited, then editorially reviewed before publishing.