This paper provides a comprehensive review of recent advancements in generative artificial intelligence(AI), focusing on two major areas: Large Language Models (LLMs) for text prediction and diffusion/GAN-based models for image generation. LLMs, such as LLaMA, BERT, GPT, Claude and O1, have revolutionized natural language processing by enabling high-quality text generation and understanding through transformer-based architectures. Simultaneously, considerable progress has been achieved with diffusion models and Generative Adversarial Networks (GANs) in generating realistic images from textual descriptions, exemplified by models like DALL \(\cdot \) E, Stable Diffusion, and StyleGAN. This review examines the underlying architectures, training methodologies, and performance benchmarks of these models. Additionally, it explores the challenges inherent in these technologies, such as resource demands, model biases, and ethical concerns. Through a comparative analysis, the paper highlights the current state of generative AI and identifies potential areas for future research.