The development phase of a compact and effective language model inspired by the LLaMA model is described in this paper. We describe our work in creating the LLM by adhering to the LLaMA principles which have guided our architectural and methodological decisions. Our approach focuses on innovation and exploration of new research avenues while the model is therefore an ongoing project. Our training relied on open-source datasets and advanced training techniques. It shows that significant progress was achieved without relying on extensive computational resources or proprietary data. Our model is still under development due to limitations in computational resources. Researchers who have access to more powerful computational resources could further and, finally, we need to refine and improve the model. The paper is aimed at motivating scholars and policymakers alike contribute to training more powerful language models. The end goal is to train improvements accessible to everyone. The models were trained on context_window, n_layers, batch_size, and d_model. And results measured on epoch, execution time, numbers of parameters, and validation loss.