LightLX — Run models too big for RAM on Apple Silicon — any Hugging Face model, dense or MoE — by streaming weights from disk

LightLX is a new tool enabling Apple Silicon users to run LLMs larger than their available RAM. It streams model weights directly from disk, supporting both dense and Mixture-of-Experts (MoE) architectures from Hugging Face. This significantly expands the accessibility of large local models for Mac users.

Opening Kapyn…