Quadrupeds and hexapods are known by their ability to adapt their locomotive patterns in function of the environment. Computational modeling of animal movement can help to better understand the emergence of locomotive patterns and their body dynamics. Although considerable progress has been made in this subject in recent years, the strengths and limitations of kinematic simulations at the scale of small moving animals are not well understood. In response to this, this work evaluated the effects of modeling uncertainties on kinematic simulations at small scale. In order to do so, a multibody model of a Messor barbarus ant was developed. The model was built from 3D scans coming from X-ray micro-computed tomography. Joint geometrical parameters were estimated from the articular surfaces of the exoskeleton. Kinematic data of a free walking ant was acquired using high-speed synchronized video cameras. Spatial coordinates of 49 virtual markers were used to run inverse kinematics simulations using OpenSim software. The sensitivity of the model's predictions to joint geometrical parameters and marker position uncertainties was evaluated by means of two Monte Carlo simulations. The developed model was four times more sensitive to perturbations on marker position than those of the joint geometrical parameters. These results are of interest for locomotion studies of small quadrupeds, octopods and other multi-legged animals.