One computer is busy. Another is sitting idle. NVIDIA’s Personal AI Router, or PAIR, is designed to send separate AI requests to compatible machines with room to handle them.
The free, open-source beta was announced around IFA on September 3. It supports Windows, macOS and Linux, including supported RTX hardware and Apple M4 or newer systems. Ollama or LM Studio still runs the model.
What the extra computer changes
The useful case is several independent tasks happening together: different agents can work without every request waiting for the same machine. PAIR assigns each request to one suitable computer; it does not combine their memory into one giant GPU.
That distinction matters when choosing a workflow. Parallel research or multiple local tools have more opportunity to benefit than a single request that must finish before the next one begins.
A concrete starting point
Pair compatible computers, enable a supported local engine and make the required models available. NVIDIA’s Jobs view shows where requests actually ran, so you can check whether the second machine contributed.
In NVIDIA’s example, a five-agent task averaged 18 minutes on one laptop and 8 minutes 48 seconds across three devices. That is a specific company demonstration, not a speed guarantee for your setup. The interesting question is whether your own waiting work can be shared.
Source published 2026-09-03. Coverage is based on the maker’s announcement and demonstration.