<p>Drug development for digestive tumors depends on various preclinical models to evaluate their efficacy and safety. Traditional models, such as 2D cell lines and animal models, fail to recapitulate the complex pathology in the human body. Recently, novel models, such as 3D organoids, tumor spheroids, and organ-on-a-chip systems, have undergone rapid growth. These models can recapitulate tissue architecture and the microenvironment in a more faithful way, enhancing the translational relevance of in vitro experiments to clinical outcomes. Moreover, patient-derived xenografts and genetically engineered models retain the genetic heterogeneity and immune contexture of original tumors, which play critical roles in assessing drug efficacy and resistance mechanisms. This review aims to explore the strengths and limitations of diverse models in drug discovery for digestive tumors. In addition, we delineate their utility in target discovery or validation, lead compound screening, and preclinical mechanistic profiling. Current studies indicate that drug failure rates can be minimized by the integration of 2D high-throughput screening, 3D organoid/tumor spheroid screening, and in vivo animal validation. Multimodal synergy and personalized models can lead to the development of more efficient and precise approaches for treating digestive tumors.&#xa0;</p> Graphical abstract <p>Created in <a href="https://BioRender.com">https://BioRender.com</a>.</p> <p></p>