Background. Sex estimation is the first step in profiling unidentified skeletal remains, and the pelvis is the most sexually dimorphic skeletal region. Pelvic osteometric standards are population-specific, and none exists for Afghanistan. Objective. To develop pelvic osteometric standards and prediction models for sex estimation in adult autopsy cases in Kabul, Afghanistan. Methods. This cross-sectional study analysed pelves of 100 adults (50 males, 50 females; 18–65 years) from unexpected or unnatural deaths examined in Kabul in 2021–2023. Pelvic width, pelvic height, superior and inferior acetabular diameters and subpubic angle were measured with a digital caliper and goniometer and analysed with Student’s t-tests, binary logistic regression and receiver operating characteristic analysis. Results. Males had larger pelvic width (28.5 ± 2.4 vs 25.2 ± 2.1 cm), pelvic height and acetabular diameters (all p < 0.001); females had a wider subpubic angle (82.3 ± 3.8° vs 75.2 ± 4.1°; p < 0.001). All five dimensions were independent predictors, pelvic width having the largest adjusted odds ratio (aOR 1.52; 95% confidence interval, CI, 1.26–1.83). The full model had an area under the curve (AUC) of 0.89 (95% CI 0.82–0.95) and classified 81.0% of cases correctly at a 0.50 cut-off (sensitivity 80.0%, specificity 82.0%; 80.0% on leave-one-out cross-validation). The subpubic angle alone reached an AUC of 0.90. Iliac and acetabular sub-models, simulated on intact specimens, retained 77.0% and 76.0% accuracy. Conclusion. Kabul-specific pelvic models correctly classify about four in five adults and offer a low-cost probabilistic tool for sex estimation, pending validation on fragmented remains.