This technical report introduces two scaled-down variants of the Stable Diffusion XL (SDXL) model, named Segmind Stable Diffusion (SSD-1B) and Segmind-Vega, which have 1.3B and 0.74B parameter UNets, respectively. The models are achieved through progressive removal of residual networks and transformer blocks from the U-Net structure of SDXL, focusing on reducing model size while preserving generative quality. The methodology involves layer-level losses to identify and retain essential features while discarding redundant elements. The compressed models effectively emulate the original SDXL by capitalizing on transferred knowledge, achieving competitive results against larger multi-billion parameter SDXL models. The work highlights the efficacy of knowledge distillation coupled with layer-level losses in reducing model size while maintaining high-quality generative capabilities, making SDXL more accessible for diverse applications in resource-constrained environments. The models are released at <https://hf.co/Segmind>.This technical report introduces two scaled-down variants of the Stable Diffusion XL (SDXL) model, named Segmind Stable Diffusion (SSD-1B) and Segmind-Vega, which have 1.3B and 0.74B parameter UNets, respectively. The models are achieved through progressive removal of residual networks and transformer blocks from the U-Net structure of SDXL, focusing on reducing model size while preserving generative quality. The methodology involves layer-level losses to identify and retain essential features while discarding redundant elements. The compressed models effectively emulate the original SDXL by capitalizing on transferred knowledge, achieving competitive results against larger multi-billion parameter SDXL models. The work highlights the efficacy of knowledge distillation coupled with layer-level losses in reducing model size while maintaining high-quality generative capabilities, making SDXL more accessible for diverse applications in resource-constrained environments. The models are released at <https://hf.co/Segmind>.