Deep learning has truly changed the game across numerous fields, reshaping how we tackle complex challenges by providing highly precise and efficient solutions tailored to particular needs. Just picture a system that can create text, summarize information, translate languages, classify data, answer questions, and even reason— deep learning makes all of this a reality. In this review, we took a closer look at different deep learning architectures and see how they drive these various applications. We analysed the past studies and reveal the datasets that power these models, as well as the design principles that influence their performance. Throughout this we emphasized the strengths that set these architectures apart, along with the limitations that pose challenges to their effectiveness. This review acts as a guide for researchers, practitioners, and industry professionals, helping them choose and adapt the right deep learning models for specific tasks