The Drag-based Model (DBM) is an analytical model for heliospheric propagation of Coronal Mass Ejections (CMEs) that predicts the CME arrival time and speed at Earth or any other given target in the solar system. It is based on the equation of motion and depends on initial CME parameters, background solar wind speed, w and the drag parameter γ. A very short computational time of DBM (< 0.01s) allowed us to develop the Drag-Based Ensemble Model (DBEM) that considers the variability of model input parameters by making an ensemble of n different input parameters to calculate the distribution and significance of the DBM results. Using such an approach, we apply DBEM to determine the most likely CME arrival times and speeds, quantify the prediction uncertainties and calculate the confidence intervals. Recently, a new DBEMv3 version was developed including the various improvements and Graduated Cylindrical Shell (GCS) option for the CME geometry input as well as the CME propagation visualizations. Thus, we compare the DBEMv3 with previous DBEM versions (e.g. DBEMv2), evaluate it and determine the DBEMv3 performance and errors by using various CME-ICME lists. Compared to the previous versions, the DBEMv3 provides very similar results for all calculated output parameters with slight improvement in the performance. Based on the evaluation performed for 146 CME-ICME pairs, the DBEMv3 performance with mean error (ME) of -11.3 h, mean absolute error (MAE) of 17.3 h was obtained, similar to previous DBM and DBEM evaluations. Fully operational DBEMv3 web application was integrated as one of the ESA Space Situational Awareness portal services (https://swe.ssa.esa.int/current-space-weather) providing an important tool for space weather forecasters.