Developer releases custom inference engine for Kimi K3

A developer created a custom inference engine enabling Kimi K3 to run on a single CPU with 8 GB of RAM. The project addresses the lack of local accessibility for the model, which typically requires multi-node GPU clusters for inference.