Add CPU-optimized quantized SLM model

initial commit