Abstract: Based on the jamming phase transition effect of granular media, the flexibility and variable stiffness characteristics of soft robotic fingers were explored in this study through numerical simulation. A finite element-discrete element (FEM-DEM) coupled model integrating the nonlinear mechanics of granular contact and the evolution of packing states was established using Abaqus software, and the dynamic responses of robotic fingers under different granular volume fractions were systematically simulated. The results indicate that the granular volume fraction is a key parameter for controlling the mechanical properties of robotic fingers: when the volume fraction is below 50%, the system remains in an ultra-loose packing state, with the stiffness increase limited (increase of less than 2%). Distinct differences are observed in the measurable performance of soft robotic fingers when the volume fraction falls within different packing state thresholds. Specifically, when the volume fraction exceeds the critical threshold (approximately 55%) and enters the random close packing state, the end displacement of robotic fingers is significantly reduced (up to 56.74%), and the overall stiffness exhibits a nonlinear jump with a maximum increase of 151.29%. Thus, the transition from a high-flexibility state to a high-stiffness state is achieved. The variable stiffness mechanism based on granular jamming is clarified in this study, which provides a theoretical basis and an effective simulation design tool for the structural optimization and performance prediction of soft robotic fingers.