ds4DeepSeek 4 Flash local inference engine for Metal and CUDA 65 824 HN pointsMIT last commit 2026-07-17apple-siliconcdeepseekggufglminference-enginellm-inferencelocal-ailocal-llmmacos SourceAdd to my stackDiscuss Share:AboutMetal-first MoE inference for Apple Silicon with bounded SSD expert streaming and local OpenAI Chat, Responses, and Anthropic Messages endpoints.LanguagesC48.5%Objective-C21%C++11.8%Metal10.1%Python6.6%Shell1.3%Makefile0.5%Jinja0.2%Contributors3023615743332222222+16FeaturesDiscussNo features listed.Comments ThemeGitHub LightGitHub Light High ContrastGitHub Light Protanopia & DeuteranopiaGitHub Light TritanopiaGitHub DarkGitHub Dark High ContrastGitHub Dark Protanopia & DeuteranopiaGitHub Dark TritanopiaGitHub Dark DimmedPreferred color schemeTransparent DarkNoBorder LightNoBorder DarkNoBorder GrayRStudio CobaltPurple DarkGruvboxGruvbox DarkGruvbox LightCatppuccin LatteCatppuccin FrappéCatppuccin MacchiatoCatppuccin MochaFroCustom (experimental)InstallBuild from sourceKeywordsapple-siliconcdeepseekggufglminference-enginellm-inferencelocal-ailocal-llmmacosmetalmixture-of-expertsmoeon-device-aiopenai-compatibleqwenssd-streaming