To quantify acetabular version using 3 D reconstructions based on biplanar radiographs (BPR) with CT as reference standard. No institutional review board approval was needed. Nine dry-bone pelvises underwent BPR in five different positions (rotation/tilt). The 3 D models of each pelvis were reconstructed by two radiologists on the basis of anatomical landmarks using semi-automated software. Automated software was used to assess the 3 D models and to calculate acetabular versions perpendicular to the anterior pelvic plane on all levels in the craniocaudal direction in 1 mm steps. Transverse CT images perpendicular to the anterior pelvic plain were reconstructed through the acetabulum in 1 mm steps. Both readers measured acetabular version on each image. Inter-reader agreement was calculated. Measurements based on BPR and CT were compared. Inter-reader agreement was almost perfect for BPR-based acetabular version measurements (ICC (intraclass correlation coefficient) = 0.920, P < 0.0005) and CT (ICC = 0.990, P < 0.0005). Correlation of acetabular versions between the five BPR-positions was substantial/almost perfect (ICC = 0.722–0.887 and 0.749–0.872 for readers 1 and 2, respectively; most P < 0.0005). The acetabular version measurements between the ap-positioning from BPR and CT showed moderate agreement (mean CCC (concordance correlation coefficient) = 0.733 for reader 1, CCC = 0.755 for reader 2). Acetabular version on multiple levels can be measured using BPR and dedicated post-processing software and is relatively independent of pelvic rotation and tilt.