Neuromuscular modeling enables biomechanical analyzes through simulation of human movement, such as the pendulum test, used to assign the level of severity of the phenomenon of spasticity. The optimization process in the SCONE software allows the exploration of movement parameters, with the minimization of cost functions available for use in its libraries. Objective: To investigate which cost functions, among those available by SCONE, are most effective for generating kinematics of a normalized pendulum test. Materials and methods: A Hill-type biomechanical model was used, with 8 muscles per leg, adapted to the pendulum test (sitting position, with trunk inclined at 45º, and right lower limb suspended). Three cost functions available in SCONE were compared: one for endurance (Cubed muscle stress) and two for metabolic energy expenditure (Wang and Uchida). The Covariance matrix adaptation Evolution Strategy algorithm was applied to the parameter relating to the threshold of the tonic stretch reflex of each muscle in the model, being selected to obtain the normalized test kinematics. Information on the trend curve, best fitness, best generation and progress was collected and compared. Results: The Uchida cost function converged faster and showed the best fitness. Conclusion: The energy expenditure cost function proposed by Uchida compared to other cost functions showed greater computing speed when equating the best combination of parameters in the 61st generation and better fitness, being a more current function with an equational solution that incorporates greater physiological precepts.