Background Small cell carcinoma of the esophagus (SCCE) is an exceptionally rare subtype of esophageal carcinoma. Accurate survival prediction is challenging due to the lack of widely recognized prognostic models. This study aimed to construct and validate a prognostic model to predict overall survival (OS) in SCCE patients.Methods A total of 491 SCCE patients were included from two sources: the Fourth Hospital of Hebei Medical University (n = 333, 2010-2020) and the SEER database (n = 158, 2000-2020). Patients were subsequently divided into training (n = 234), internal validation (n = 99), and external validation cohorts (n = 158). A prognostic model for OS was constructed using multivariable Cox regression in the training cohort, from which a relative survival risk score and nomogram were derived. Model performance was evaluated using the C-index, AUROCs, calibration curves, and decision curve analysis (DCA), and compared to TNM and VASLG staging systems. The Kaplan-Meier method estimated survival, and differences were assessed using the log-rank test.Results Of the 491 patients, 314 (86.7%) were male, with a mean age of 66 years. Independent prognostic factors for OS, including TNM stage, surgery, and chemotherapy, were incorporated into a Cox model, termed the TSC model. The C-index for the TSC score in the training cohort (0.738; 95% CI, 0.615-0.845) was significantly higher than TNM (0.706; 95% CI, 0.507-0.796) and VASLG (0.657; 95% CI, 0.606-0.708). Likewise, AUROCs for the TSC score at 1, 3, and 5 years (0.713, 0.732, 0.816) outperformed both TNM (0.686, 0.682, 0.725) and VASLG (0.592, 0.609, 0.648). Moreover, calibration curves illustrated strong alignment between predicted and observed survival probabilities. DCA showed the nomogram provided superior net clinical benefits. High-risk patients had a median OS of 9.7 months, significantly shorter than 28.5 months for low-risk patients. These findings were validated in internal and external cohorts.Conclusions To the best of our knowledge, the TSC model is the first fully validated prognostic model for SCCE, offering more accurate OS predictions than TNM and VASLG staging systems, and providing a valuable tool for personalized treatment.