The design of actuators for grippers of humanoid robots offers many challenges such as withstanding significant loads, while considering anthropomorphic design aspects. Therefore, clarity is lacking about the ideal design path for the components of the actuator of a humanoid robotic gripper. This research presents a combined modeling and experimental investigation of actuator components, aimed at the development of a highly stiff and compact actuator for the gripper of humanoid robot LOLA. To determine the required grasping and holding torque of the gripper, a comprehensive Simulink model of the gripper’s finger was developed, incorporating realistic physical properties, such as friction modeling and stiffness characteristics. Besides, Simscape models were developed to capture the characteristics of various drivetrain components. Two experimental test benches, being a gripper test bench and a prototype test bench, were designed to characterize and validate the models and evaluate performance parameters under static and dynamic conditions. Three drivetrain components were experimentally analyzed: a worm gear, a planetary gear, and a roller clutch. The analysis identified holding torque, input torque, stiffness, efficiency, integration feasibility, weight, and volume as the most relevant metrics for component evaluation. In contrast, parameters such as cost, backlash, and total actuator weight were found less suitable for a meaningful analysis due to dependency on specific motor choices and lack of cost data for custom components. Contrary to theoretical expectations, the worm gear exhibited backdriving under load, while the planetary gear showed predictable backdriving behavior and limited torque capacity. The custom designed roller clutch effectively prevented backdriving up to its rated load and offered the highest torsional stiffness among the tested solutions. Although a roller clutch therefore seems to be a suitable drivetrain component of the actuator solution for the LOLA gripper, the necessary integration of a planetary gear increases the total weight and volume. Therefore, the research concludes that a non-backdrivable worm gear represents the most suitable component solution, balancing high holding torque, stiffness, and integration feasibility. However, due to limitations such as non real-time control of the prototype test bench and large roller clutch dimensions, several suggestions for further research are made. First, it is suggested to enhance the prototype test bench with real time control to enable more complex experiments. Moreover, the development of a more accurate roller clutch Simscape model is advised that better captures its dynamics. By considering these suggestions, the developed modeling and testing framework could further enhance the exploration of conventional and more complex drivetrain component configurations for an actuator tailored to humanoid robotic grippers.