ds4

ds4 is an inference engine optimized for running DeepSeek 4 Flash models locally on Apple Metal and NVIDIA CUDA hardware. It is designed for developers and AI enthusiasts who need fast, private, offline inference without cloud dependencies. Its interest lies in enabling high-performance local execution of a cutting-edge model across both major GPU ecosystems, bridging accessibility and efficiency.