Esophageal cancer (EC) is the tenth most common cancer worldwide and has high morbidity and mortality. Its main subtypes include esophageal squamous cell carcinoma and esophageal adenocarcinoma, which are usually diagnosed during their advanced stages. The biological defects and inability of preclinical models to summarize completely the etiology of multiple factors, the complexity of the tumor microenvironment, and the genetic heterogeneity of tumors severely limit the clinical treatment of EC. Patient-derived models of EC not only retain the tissue structure, cell morphology, and differentiation characteristics of the original tumor, they also retain tumor heterogeneity. Therefore, compared with other preclinical models, they can better predict the efficacy of candidate drugs, explore novel biomarkers, combine with clinical trials, and effectively improve patient prognosis. This review discusses the methods and animals used to establish patient-derived models and genetically engineered mouse models, especially patient-derived xenograft models. It also discusses their advantages, applications, and limitations as preclinical experimental research tools to provide an important reference for the precise personalized treatment of EC and improve the prognosis of patients.