The free, open-source tool sends separate Ollama and LM Studio tasks to trusted computers on one local network. It does not combine their memory into one larger GPU.
New to this? Read it in simple words
- Nvidia’s free PAIR tool sends AI tasks to several computers on one local network. Each task still runs on one computer.
- In Nvidia’s own test, three computers finished five tasks in 8 minutes and 48 seconds. One RTX Spark laptop needed 18 minutes.
- PAIR can make local AI faster. But users must decide which computers and networks they can trust.
- Users may get different results with other models, networks or machines.
- Local network
- A group of computers connected in one place, like a home or office.
- GPU
- A chip that does many calculations at once. Most AI runs on GPUs.
What PAIR does
PAIR stands for Personal AI Router. It finds trusted computers on the same network and checks which AI models they can run. It then sends each separate request to one suitable computer. Apps can keep using familiar connections for Ollama or LM Studio.
The tool is useful when one AI task creates many smaller jobs. Several jobs can run on different computers instead of waiting for one busy graphics processor. However, one large model must still fit in the memory of the computer that runs it.
PAIR sends each Ollama or LM Studio request to one trusted local computer through an encrypted connection.
Which computers can join
The beta version supports Windows, Linux, and macOS. Nvidia lists many recent RTX computers, RTX PRO systems, DGX Spark, and Apple computers with M4 or newer chips. Different types of computers can join the same group if they can run a supported model.
Nvidia tested five AI helper tasks. Three computers finished the work in 8 minutes and 48 seconds. One RTX Spark laptop needed 18 minutes. This was Nvidia’s own test, so users may get different results with other models, networks, or machines.
Local does not always mean private
PAIR uses encrypted connections between approved computers. Local apps connect through the same machine. Nvidia warns that the short setup PIN is not strong protection by itself, so people should pair devices only on a network they trust.
Users must also know who controls each computer and where logs are saved. Some apps or models may still contact online services. PAIR is a useful traffic system for local AI, but privacy depends on the full setup, not only on the word “local.”
Sources
Every fact in this story comes from the sources below. Open them to check our work.
- 1Primary source · September 3, 2026NVIDIA PAIR Virtual Inference Router Expands Available Compute on Your Local Network Nvidia Developer Blog
- 2
- 3Research · September 3, 2026Nvidia PAIR utility joins every GPU in your home into a cluster for agentic AI tasks Tom’s Hardware
We checked Nvidia’s technical article, its open-source code page, and an independent hardware report. We clearly separate Nvidia’s test result from a general speed promise. We also explain that PAIR routes separate jobs and does not combine computer memory.