NVIDIA at IFA 2026: llama.cpp up to 1.9x on RTX 5090 and PAIR to spread inference across the LAN
In one sentence On September 3 at IFA NVIDIA announced local-AI optimizations (llama.cpp up to 1.9x on RTX 5090, vLLM up to 1.4x on two DGX Spark), one-click local setup for Hermes Agent and OpenClaw, and PAIR, an open-source router that spreads inference requests across home PCs.
At the IFA show in Berlin NVIDIA presented a package of updates for running AI directly on your own PC, without going through the cloud.
The first is speed: the most common tools for running models locally, such as llama.cpp (which sits under LM Studio and Ollama), run up to almost twice as fast on RTX 5090 cards. The second is simplicity: agents like Hermes Agent and OpenClaw now set up a suitable local model for the installed graphics card on their own.
The third is PAIR, a free open-source tool that uses several computers on the same network as one: if a PC is busy, the request goes to the one with a free GPU. NVIDIA also announced RTX Spark Windows PCs, arriving in October.
Companies
NVIDIA, Nous Research, Perplexity, Lenovo, Acer
Tools
llama.cpp, vLLM, LM Studio, Ollama, NVIDIA PAIR, Hermes Agent, OpenClaw
Tags
Sources