Introduction: Rapid bedside risk stratification of burn-injured patients informs intensive-care admission, fluid-resuscitation intensity, and timing of early surgical excision and grafting. The Modified Abbreviated Burn Severity Index (mABSI) and the Revised Baux (R-Baux) score are widely cited but have not previously been compared head-to-head in a Sumatran tertiary burn unit. Methods: STARD-aligned retrospective diagnostic-accuracy study of consecutive thermal-burn patients admitted to the Burn Unit, Dr. Mohammad Hoesin General Hospital Palembang (January–December 2025). Patients with concurrent major trauma or incomplete records were excluded. Both scores were calculated from medical records by an investigator blinded to the outcome. In-hospital mortality was the reference standard. Receiver-operating-characteristic, Wilson-CI sensitivity/specificity, DeLong AUC comparison, McNemar paired test, and Cohen's κ analyses were performed in SPSS v25 (α = 0.05). Results: Of 45 included patients (mean age 40.8 years; 77.8% male; 64.4% fire burns), 20 (44.4%) died. Optimal cutoffs were mABSI ≥ 5 (AUC 0.811, 95% CI 0.688–0.934; sensitivity 72.0%, specificity 75.0%, Youden 0.47) and R-Baux > 60 (AUC 0.808, 95% CI 0.678–0.938; sensitivity 52.0%, specificity 82.0%, Youden 0.34). Inter-test agreement was substantial (κ = 0.69, 95% CI 0.48–0.90; p < 0.001), with overall accuracy of 84.4% when R-Baux was treated as the comparator. Conclusion: Both scores discriminated burn mortality well; the Modified ABSI offered higher sensitivity and Youden index, supporting its evaluation as a primary triage tool in similar Indonesian tertiary burn units, while the Revised Baux retained complementary specificity for confirmatory risk classification.