DwarfStar 4 (ds4) - Local LLM Inference Engine

CLICK TO OPEN

A C-based inference engine for running frontier open-weight models (DeepSeek V4/V4.1, Qwen, GLM) locally on high-memory machines. Supports Metal, CUDA, and ROCm with CLI, HTTP APIs, and native agent capabilities. MIT-licensed open-source project by antirez (Redis creator).