NVIDIA's PAIR Beta Enables Distributed AI Workloads Across Local Networks

NVIDIA has launched a beta version of its new tool, PAIR (Platform for AI in Real-time), designed to enhance local AI development by distributing inference requests across compatible personal computers on a network. This open-source solution automatically identifies available PCs and routes tasks to systems with available capacity, aiming to streamline complex local agent workflows that involve breaking down tasks into smaller, manageable jobs.

PAIR supports major operating systems including Windows, macOS, and Linux, and is compatible with a range of NVIDIA GPUs starting from the RTX 20-series, as well as workstation GPUs and Apple's M4 chips and newer. The tool integrates with popular AI platforms like Ollama and LM Studio. NVIDIA also reports significant performance improvements in its latest IFA updates, including enhanced local model setup for agents like Hermes, OpenClaw, and Perplexity Portable Computer, and up to 1.9x higher throughput for llama.cpp on their RTX 5090 GPU.