Floods represent one of the most destructive natural hazards, posing significant risks to both ecological systems and human societies. Consequently, the development of accurate and efficient predictive and management models has become a critical research priority. This paper presents a review of the historical progression and technological advancements in flood modeling, encompassing traditional empirical and conceptual models, physics-based hydrodynamic models derived from the Saint-Venant (SV) and Shallow Water Equations (SWEs), and contemporary approaches powered by Artificial Intelligence (AI), Machine Learning (ML), and Deep Learning (DL). The review critically examines the strengths and limitations of these methodologies: hydrodynamic models achieve high physical realism but require intensive computational resources and detailed geospatial data, whereas AI-based models offer remarkable computational efficiency and scalability in handling large datasets but often struggle with issues of interpretability, generalization, and physical consistency. The paper also highlights the emerging paradigm of hybrid modeling frameworks that integrate physical principles with AI algorithms, such as Physics-Informed Neural Networks (PINNs) and Graph Neural Networks (GNNs), which have shown promise in enhancing both predictive accuracy and computational performance. Moreover, it identifies ongoing challenges, including data scarcity, limited model transferability, and the pressing demand for explainable AI (XAI). The study concludes that the future of flood modeling lies in the synergistic integration of physics-based and data-driven approaches, supported by advances in remote sensing, cloud computing, and the Internet of Things (IoT), to build more robust and efficient early warning systems capable of addressing the intensifying impacts of climate change.