Computational models have been seminal in expanding our understanding of how neural, biomechanical, or physical elements can give rise to behavior. With the forthcoming emergence of nearly complete connectomes of the central nervous system, musculature, and biomechanics of Drosophila, it is only now possible to construct models grounded in anatomy and physiology. Biological experiments, in reciprocity, could complement these models, verifying and enhancing their accuracy. Here, we leverage a biomechanical model of Drosophila to build, for the first time, a Hill-type muscle model using X-ray scans of Drosophila to recapitulate limb kinematics during multiple behaviors. Overall, we modeled 13 out of ~20 muscles in the fly's front leg, spanning three joints in OpenSim. The anatomical properties of the muscles---the attachment points and the muscle-tendon length---were based on the X-ray scans from two animals. Next, we built an optimization pipeline to identify the unknown muscle parameters. Since little is known about the physiology of the fly leg muscles (e.g., maximum isometric force), such parameters were initialized based on the literature on other fly muscles. We optimized the muscle parameters of the legs to replicate the detailed inverse kinematics computed from the 3D pose of the fly during antennal grooming and front leg rubbing. We analyzed the muscle moment arms, induced torque, and derived muscle activation to understand the relative contribution of each muscle to the observed motion. Finally, to gain computational speed and perform contact-rich simulations, we transferred our OpenSim model into the physics engine MuJoCo and validated the conversion. In conclusion, our model can assist biological experiments in elucidating the neural control of muscles underlying complex kinematics in a multi-degree-of-freedom system. Moreover, it can be used for actuating artificial agents that exhibit natural and compliant movement in simulations.