Tumor models are conventional methods for developing anti-cancer drugs, evaluating drug delivery, or calculating drug efficacy. However, traditional cell line-derived tumor models are unable to capture the tumor heterogeneity in patients or mimic the interaction between tumors and their surroundings. Recently emerging patient-derived preclinical cancer models, including of patient-derived xenograft (PDX) model, circulating tumor cell (CTC)-derived model, and tumor organoids-on-chips, are promising in personalized drug therapy by recapitulating the complexities and personalities of tumors and surroundings. These patient-derived models have demonstrated potential advantages in satisfying the rigorous demands of specificity, accuracy, and efficiency necessary for personalized drug therapy. However, the selection of suitable models is depending on the specific therapeutic requirements dictated by cancer types, progressions, or the assay scale. As an example, PDX models show remarkable advantages to reconstruct solid tumors in vitro to understand drug delivery and metabolism. Similarly, CTC-derived models provide a sensitive platform for drug testing in advanced-stage patients, while also facilitating the development of drugs aimed at suppressing tumor metastasis. Meanwhile, the demand for large-scale testing has promoted the development of tumor organoids-on-chips, which serves as an optimal tool for high-throughput drug screening. This review summarizes the establishment and development of PDX, CTC-derived models, and tumor organoids-on-chips and addresses their distinctive advantages in drug discovery, sensitive testing, and screening, which demonstrate the potential to aid in the selection of suitable models for fundamental cancer research and clinical trials, and further developing the personalized drug therapy.