The article discusses a novel approach for optimizing atomic-molecular structures in extra-large clusters, focusing on reducing computational complexity and applying parallel computing techniques. A three-stage computation process is outlined, offering benefits over traditional algorithms by efficiently navigating through multi-extremal optimization challenges. The software, based on this methodology, demonstrated high efficacy in solving complex molecular dynamics problems. When comparing the outcomes with experimental data, no significant deviations from the expected patterns in potential values were detected. The developed three-stage numerical approach proved instrumental in solving optimization tasks for extra-large atomic-molecular systems.