Objective To establish the musculoskeletal multi-body dynamic foot-ground contact model and explore its applicability at different speed. Methods The gait data of the subjects at different speed were collected, and the foot-ground contact model was established based on the full body model from the musculoskeletal multibody dynamic software AnyBody. Then the calculated ground reaction forces (GRFs) and ground reaction moments (GRMs) at different speed (slow walking, normal walking, fast walking and jogging) were compared with the measurements from the force plates. Results The predicted GRFs and GRMs correlated well with the experimental measurements at slow, normal and fast speed (stride speed ranged from 0.69 to 1.68 m/s). The correlation coefficients between predicted and measured GRFs were greater than 0.875 and the correlation coefficients for GRMs were greater than 0.9. Conclusions The developed foot-ground contact model could simultaneously predict GRFs and GRMs with good accuracy, thus eliminating the dependency on force plates. The model could be applied to low-speed gait conditions, such as the elderly and pathological gait.