A novel stochastic algorithm for global optimization, Molecular-Inspired Parallel Tempering (MIPT), is presented. MIPT incorporates some basic features of molecular dynamics simulation into the Parallel Tempering formulation. In MIPT, molecules move in the decision-variable-space as the result of different forces: repulsion, friction and random forces. Two different types of molecules are considered: explorers and refiners. Explorers present lower friction and are subject to repulsion forces causing them to move faster towards low molecular density regions. Refiner molecules achieve better values of the objective function and are subject to larger friction forces restricting their motion to a narrow region around their current position. The efficiency of MIPT is tested in five challenging case studies and compared with other established, well-known optimization methods. The results demonstrate that new MIPT is a competitive and efficient algorithm, reaching the global optimum with 100% success ratio in most cases, without requiring much computational effort. (C) 2010 Elsevier Ltd. All rights reserved.