Nvidia Releases PAIR Beta to Distribute AI Agent Tasks
Nvidia released the open-source PAIR beta system to accelerate AI agents by offloading parallel tasks to other computers on a local network.
Nvidia Inc. released the beta version of PAIR, or Personal AI Router, an open-source system that accelerates AI agents by offloading parallel tasks to other computers on a local network. The software prevents a single system's memory pool from becoming crowded during GPU-intensive tasks by allowing a primary system to distribute subagents to other Windows, Mac, or Linux machines.
PAIR utilizes mDNS for device discovery and MTLS for security. It is currently compatible with the LM Studio and Ollama engines. To operate the system, the primary machine requires an RTX-class GPU, while remote Mac systems must be equipped with an M4-generation processor or later.
The system does not pool GPU processing or memory to run larger models. Instead, it assigns individual subagents to separate systems based on available bandwidth and current workload.
Keep reading in the app
The full story and every source, free in the app.
Coming soonGoogle Play