The effectiveness of the plantarflexor muscle group to generate desired plantarflexion moments is modulated by the geometry of the Achilles tendon moment arm (ATMA). Children with cerebral palsy (CP) frequently have reduced plantarflexion function, which is commonly attributed to impaired muscle structure and function, however little attention has been paid to the potential contribution of ATMA geometry. The use of musculoskeletal modelling for the simulation of gait and understanding of gait mechanics, rely on accuracy of ATMA estimates. This study aimed to compare 3D in-vivo estimates of ATMA of adults, children with CP and typically developing (TD) children, as well as compare 3D in-vivo estimates to linearly scaled musculoskeletal model estimates. MRI scans for eight children with CP, 11 TD children and nine healthy adults were used to estimate in-vivo 3D ATMA using a validated method. A lower limb musculoskeletal model was linearly scaled to individual tibia length to provide a scaled ATMA estimate. Normalised in-vivo 3D ATMA for children with CP was 17.2% +/- 2.0 tibia length, which was significantly larger than for TD children (15.2% +/- 1.2, p = 0.013) and adults (12.5% +/- 0.8, p <0.001). Scaled ATMA estimates from musculoskeletal models significantly underestimated in-vivo estimates for all groups, by up to 34.7%. The results of this study show children with CP have larger normalised 3D ATMA compared to their TD counterparts, which may have implications in understanding reduced plantarflexor function and the efficacy of surgical interventions whose aim is to modify the musculoskeletal geometry of this muscle group. (C) 2018 Published by Elsevier Ltd.