This work addresses a balancing problem approach whose objective is to minimizing the ergonomic risk dispersion between the set of workstations of a mixed-model assembly line and the risk level of the workstation with the greatest hazard. A GRASP procedure is proposed to achieve these two challenges simultaneously. This new procedure is compared with two Mixed Integer Linear Programs, the MILP-1 that minimizes the maximum ergonomic risk of the assembly line, and the MILP-2 that minimizes the average deviation from ergonomic risks of the set of workstations of the line. The results from the case study linked with the automotive sector indicate that the proposed GRASP is a very competitive and promising tool for further researches.