In this study, researchers developed a machine learning–based model to predict long-term survival outcomes in patients with spinal chordoma and chondrosarcoma. A total of 3,175 patients were analyzed, including 1,971 with chondrosarcoma and 1,204 with chordoma. The primary outcome was 10-year survival, while secondary outcomes included 30-day readmission rates and short- and mid-term mortality at 30 days, 90 days, 1 year, and 5 years. Twelve machine learning algorithms were evaluated, including Gradient Boosting, CatBoost, XGB, and deep learning models such as CNN, GRU, LSTM, SimpleNN, and DeepSurv. Model performance was assessed using Receiver Operating Characteristic (ROC) curves, Area Under the Curve (AUC), Brier Scores, and the Concordance Index (C-index). The dataset was split into training (80%) and testing (20%) subsets for validation. Overall, the 10-year mortality rate was 48.1% for chordoma and 54.7% for chondrosarcoma. Key predictors of long-term survival included age, tumor type, radiation therapy, and insurance status. Among the models tested, the Gradient Boosting Classifier achieved the highest accuracy (0.6929), while the Gaussian Naive Bayes model performed the weakest (0.485). The LightGBM (LGBM) classifier demonstrated the greatest predictive power (AUC = 0.828), whereas the SVC model had the weakest performance (AUC = 0.604). To enhance clinical translation, the researchers developed a web-based survival prediction tool for use by physicians and patients. While promising for supporting individualized clinical decision-making, external validation is necessary to confirm its generalizability.