Recently, Large Language Models (LLMs) have showcased their versatility and effectiveness not only in natural language processing but also in a broad spectrum of other fields. These applications include automated content generation, real-time language translation, sentiment analysis, personalized chatbot responses, and even assisting in data analysis and predictive modeling across various industries. This impressive showcase of abilities from LLMs has sparked an increase in research activity within this area, covering a wide range of subjects. These research efforts include improvements in neural network designs, extensions in context length, better alignment of models, enhancements in training datasets, benchmarking activities, and strides in efficiency. In our study, we tackle two primary research questions: First, what are the current abilities and limitations of Large Language Models (LLMs), and second, how are these models influencing various sectors? To answer these, we conduct a systematic literature review of peer-reviewed articles and industry reports, allowing us to comprehensively assess the breadth of applications and advancements in LLMs. Our focus centers on recent enhancements in neural network architectures, how these models process context, improvements in training datasets, and strides made in boosting model efficiency and alignment. Our paper provides a current and detailed review of the literature by delving into the historical context, significant findings, and dominant methods in the field. By analyzing various large language models, this study not only delivers a thorough overview but also maps out the existing challenges and suggests directions for future research. This survey offers a complete view of the state of generative AI today, highlighting areas for continued research, improvement, and innovative development.