This is the models of our paper "EfficientLLM: Scalable Pruning-Aware Pretraining for Architecture-Agnostic Edge Language Models".