Several patient-derived tumor models have emerged recently. However, soft tissue sarcomas (STSs) present a challenge in developing preclinical drug-testing models due to their non-epithelial and complex nature. Here, we report a model termed patient-derived tumor-like cell clusters (PTCs) derived from STS patients. PTCs result from the self-assembly and proliferation of mesenchymal stem cells (MSCs), epithelial cells, and immune cells, faithfully recapitulating the morphology and function of the original tumors. Through standardized culture and drug-response assessment protocols, PTCs facilitate personalized drug testing, evaluating hundreds of therapies within two weeks. Notably, PTCs exhibit 100% accuracy in distinguishing between complete or partial response and disease progression. We demonstrate the utility of PTCs in guiding chemotherapy selection for a patient with relapse and metastases following conventional therapy, who exhibited a positive response after non-conventional therapy identified through PTC. These findings underscore the potential of PTCs for prospective use in clinical decision-making regarding therapy selection.