NVIDIA Personal AI Router (PAIR) is beta software that routes separate local AI requests to compatible computers on the same network. It gives supported apps a common endpoint, but it does not combine GPUs, pool their memory, or split one model or request across machines. To handle a request, one eligible computer must be online, running a supported inference engine, and have the requested model available.
What NVIDIA PAIR does
PAIR is a routing and coordination layer for local inference—not an AI model or a replacement for an inference engine. NVIDIA describes it as a way for AI apps and agent workflows to use compute across compatible systems on a local network. Its product page highlights machine discovery, inference routing, and setup without special cables or racks (NVIDIA PAIR product page).
| # | Preview | Product | Price | |
|---|---|---|---|---|
| 1 |
|
MINISFORUM MS-02 Ultra Workstation Mini PC, Intel Core Ultra 9 285HX (24C/24T, up to 5.5GHz), PCIe... | $1,659.00 | Buy on Amazon |
| 2 |
|
GMKtec EVO-X2 AI Mini PC Ryzen Al Max+ 395 Superchip 128GB LPDDR5X 2TB SSD | $3,649.99 | Buy on Amazon |
As an Amazon Associate I earn from qualifying purchases.
PAIR routes each request to one node that can serve it. That can help distribute independent work—for example, separate requests from concurrent agents—among eligible computers. It does not make several modest GPUs behave like one larger GPU. NVIDIA’s Getting Started guide and architecture documentation distinguish request routing from pooling GPU memory or splitting a model or request across systems.
Recommended Free Tools
How PAIR routes a request
- Discover and pair computers. Install PAIR on the participating devices and connect them on the same local network. The setup guide describes discovering systems or adding an IP address, then accepting an invitation with a temporary PIN.
- Make a node ready to infer. A computer needs a supported inference engine running and the requested model available to that engine. PAIR can install or manage an engine, or use an existing installation.
- Connect the app to PAIR. Configure a compatible AI application or agent to use PAIR’s local endpoint rather than configuring it separately for each machine.
- Send independent requests. PAIR selects an eligible node for each request. A node that is offline, lacks a running compatible engine, or does not have the requested model available cannot serve that request.
Models are associated with engines on individual computers; PAIR does not automatically share a model between nodes. If two nodes need to serve the same model, make it available on both. The engine management documentation covers supported engine setup and lifecycle.
#1 Best Overall
- High-Performance AI Processor：The MS-02 Ultra features an Intel Core Ultra 9 285HX (24C/24T, up to 5.5 GHz, 13 TOPS NPU), delivering fast and efficient performance for AI inference, algorithm development, and media workloads. A PCIe x16 expansion slot supports desktop-class GPU upgrades for advanced model training and accelerated computing tasks. It's ideal for creators, engineers, and teams handling intensive parallel workloads.
- 4 × M.2 PCIe 4.0 + 4 × DDR5 SODIMM slots:Four DDR5 SODIMM slots support up to 256 GB of memory, while ECC helps maintain data integrity in mission-critical environments. Four PCIe 4.0 M.2 slots support up to 24 TB of storage, supporting RAID 0/1/5/10, combining high-speed performance with data protection. It allows for the creation of independent scratch disks, media libraries, and project drives, providing high-throughput for production workflows.
- PCIe & USB 4.0 v2: Up to three PCIe slots can be equipped, including a dual-slot x16 GPU. The main slot supports PCIe 5.0, meeting the needs of high-bandwidth creative and computing workloads. USB 4.0 v2 (80Gbps) supports high-bandwidth external storage and displays.
- Ultra-fast Networking: Wi-Fi 7 further enhances wireless performance with next-generation speeds and low-latency stability. Intelligent bandwidth switching optimizes throughput in different network environments, ensuring optimal performance for enterprise or local networks. Dual 25GbE ports (providing up to approximately 3.125 GB/s bandwidth, about 25 times faster than traditional 1GbE), enabling seamless large-scale file transfers and parallel computing. 10GbE and 2.5GbE ports, with support for Intel vPro technology, ensure enterprise-grade remote management and deployment flexibility.
- Server-grade thermal architecture: Utilizing a dedicated CPU/GPU airflow design, equipped with a 6-pipe dual-fan cooler, it maintains stable performance even under sustained loads, delivering up to 140W Turbo power while maintaining a 100W TDP, and operating with noise levels as low as 36 dB. An integrated 350W power supply ensures stable and reliable output for demanding computing tasks and fully loaded extended configurations.
Supported engines and practical requirements
NVIDIA identifies Ollama and LM Studio as supported local inference engines. Compatibility still depends on the particular application, engine, model, operating system, and machine; the presence of PAIR alone does not make every model work on every node. Check NVIDIA’s current PAIR FAQ and project repository for current compatibility details.
| Item | NVIDIA product-page listing |
|---|---|
| Operating systems | Windows 11, DGX OS, Ubuntu 14.04, and macOS Tahoe |
| GPU or device | GeForce RTX 20 Series or newer, DGX Spark/GB10, or Mac M4 or newer |
| Memory | 8 GB or higher |
| Disk space | 20 GB or higher recommended |
| Internet | Not required for operation; required to download models |
These are configuration details listed on NVIDIA’s product page, not a guarantee that every engine and model runs on every listed system. There is also a platform-listing discrepancy worth checking before installation: the product page names Ubuntu 14.04, while NVIDIA’s repository describes Windows 11, Linux, and macOS, with x64 and arm64 support and Windows on ARM marked experimental. Consult the live download and requirements information for your exact operating system rather than assuming either listing settles compatibility.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What you need to try it
- A compatible computer for each participating node. A single machine can run local inference; multiple machines are needed to try pairing and routing.
- PAIR installed on each participating device, with the computers connected to the same local network.
- A supported engine—Ollama or LM Studio—and the model needed for the workload available on whichever node is expected to serve it.
- A compatible application or agent configured to use PAIR’s local endpoint.
For an initial setup, start with one computer and confirm that its engine and model work locally. Then add another node, pair the machines, and make the required model available on that node before directing requests through PAIR. NVIDIA’s setup guide advises pairing only devices and networks you trust.
The Tool Desk
Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Rank #2
- EVOLUTION RYZEN AI MAX+ 395 MINI PC - GMKtec EVO-X2 is the next evolution in AI mini PC Ryzen Strix Halo series. Thanks to AMD Simultaneous Multithreading (SMT) the core-count is effectively doubled, to 32 threads. Ryzen AI Max+ 395 has 64 MB of L3 cache and can boost up to 5.1 GHz, depending on the workload. The Ryzen AI Max+ 395 is currently rated as the "most powerful x86 APU" on the market for AI computing.
- AI NPU with XDNA 2 ARCHITECTURE - Powered by 16 “Zen 5” CPU cores, 50+ peak AI TOPS XDNA 2 NPU and a truly massive integrated GPU driven by 40 AMD RDNA 3.5 CUs, the Ryzen AI MAX+ 395 is a transformative upgrade and delivers a significant performance boost over the competition. The Ryzen AI Max+ 395 excels in consumer AI workloads like the llama.cpp-powered application: LM Studio. Shaping up to be the must-have app for client LLM workloads, LM Studio allows users to locally run the latest language model without any technical knowledge required and unleash their creativity and productivity.
- AMD RADEON 8090S iGPU GAMING PC - The AMD Radeon RX 8060S offers all 40 CUs with up to 2.9 GHz graphics clock and uses the new RDNA 3.5 architecture. The powerful iGPU is positioned between an RTX 4060 and 4070 laptop GPU and therefore enables gaming in FHD at maximum details in most demanding games. The 8060S can also utilize the full 128GB pool, which is perfect for running LLMs such as Deepseek 70B Q8, which runs comfortably on this machine.
- EIGHT CHANNEL LPDDR5X - LPDDR5X is a new ground breaking memory small form factor installed on-board. With blazing speeds up to to 8000MT/s, it runs 1.5x faster than the DDR5 SODIMMs; 90% better performance over DDR5 SODIMMs in video conferencing and photo editing; 30% better performance in productivity apps; 12% better performance in digital content workloads.
- QUAD SCREEN 8K DISPLAY SUPPORT - EVO-X2 AI Mini PC support 4-screen 4K/8K output via HDMI 2.1 (8K@60Hz), DisplayPort 1.4 (4K@60Hz), and dual USB 4 40Gbps Transfer speed (supporting PD3.0/DP1.4/DATA). Ideal for gaming, video editing, and multitasking, it provides expansive and crisp multi-display support.
Privacy and network boundaries
NVIDIA positions PAIR for local inference: in a correctly configured workflow, prompts, files, and agent context are intended to stay on the local network rather than being sent to a cloud inference service. That describes the intended routing, not a blanket security guarantee. The behavior of the application, engine, model-download process, participating computers, and network configuration also matters. Only pair computers on a network you trust.
When PAIR is—and is not—a fit
PAIR may be useful when you have compatible computers on one local network and want to route separate requests to nodes that each have the engine and model needed to answer them. It is not the answer to a single model that cannot fit in one computer’s memory: PAIR does not pool VRAM or divide that model’s execution across machines.
There is no independent performance benchmark established here, so routing more requests across nodes should not be read as a promised speedup for any particular workload. The practical result depends on the available nodes, their engines and models, and the requests being sent.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
