Objective To evaluate the diagnostic accuracy and reliability of the Ovarian-Adnexal Reporting and Data System (O-RADS) ultrasound v2019 in classifying adnexal masses (AMs) and compare the old and updated systems (v2022). Patients and methods This prospective study enrolled 977 consecutive women with suspected AMs from three institutions between January 2022 and December 2023. Ultrasound examinations were performed by three experienced radiologists who categorized AMs according to O-RADS ultrasound v2019. The same radiologists retrospectively reviewed the stored ultrasound images and provided the O-RADS ultrasound v2022 classification. Histopathology was used as the reference standard to calculate the diagnostic accuracy of the O-RADS versions in predicting malignant AMs. Inter-observer agreement (IOA) of the O-RADS scoring results was evaluated using the Fleiss kappa (κ) test. Results The final analysis included 803 women with 855 AMs (219 (25.6%) malignant and 636 (74.4%) benign). Both O-RADS versions demonstrated good diagnostic accuracy, with area under the curve (AUC) values ranging from 0.906 to 0.923 (v2019) and 0.919 to 0.936 (v2022). The updated v2022 showed a slightly higher accuracy (82.5–86.7% vs. 80.7–85.3%), sensitivity (93.6–95.0% vs. 92.2–94.1%), and specificity (78.1–84.1% vs. 76.1–82.9%) compared to v2019. The IOA for the overall O-RADS classification was perfect for both versions (κ = 0.96–0.97). Conclusions The O-RADS ultrasound classification system demonstrated good diagnostic accuracy and reliability in predicting malignant AMs, with the updated v2022 showing modest improvements.