Running large language models locally has always been a trade-off: either you invest in expensive hardware with 24GB+ VRAM, or your system hits an Out-Of-Memory (OOM) crash the moment a model exceeds physical RAM.
We ve been working on solving this resource wall.
This lets developers and power users run models larger than their physical RAM capacity on everyday Windows laptops without needing a dedicated GPU or suffering OOM crashes.
Laghu streams local AI models of any parameter size in as little as 1 GB RAM—turning your Windows PC into private file and tool agents without a GPU or OOM crashes.