The Framingham Heart Study (FHS) dataset includes upper-body musculoskeletal models of 125 women and 125 men, each including thoracolumbar spine, ribcage, head and neck body and upper extremities (Mokhtarzadeh et al., 2021). The intervertebral joints (from L5/S1 to T1/T2) present three rotational degrees of freedom and the costovertebral joints are modelled as pin joints. The sacrum is connected to the ground via a pin joint, only allowing for anterior and posterior tilting of the pelvis. These models include muscle groups modelled by multiple fascicles and each one of them is represented by a Hill-Type model, using Thelen2003Muscle as muscle model in OpenSim.We updated the described dataset to ensure compatibility with OpenSim 4.x. For this purpose, it is important to note that OpenSim 4.x modifies the handling of the Thelen muscle model, which now requires a minimum activation of 0.01 to avoid numerical singularities at zero activation. By contrast, the Millard muscle model (i.e. Millard2012EquilibriumMuscle) inherently prevents such singularities when activation is zero. Since OpenSim only permits updating a .osim file to the latest version without altering the muscle model, it was necessary to identify a generic MSK model using the Millard formulation in OpenSim 4.x, into which the subject-specific characteristics of each dataset model could be incorporated.The generic full-body model chosen for use in OpenSim 4.x was developed by Burkhart (Burkhart et al., 2020) and was modified to have FHS models’ features. Specifically:• Only the upper-body was maintained, eliminating the lower extremities and the associated muscles and joints.• All joint coordinates were defined as “clamped”.• Costovertebral joints were locked.• Five degrees of freedom were added to the joint connecting the model to the ground.• The range of motion of the intervertebral joints were reduced to literature values.• The pelvis COM was repositioned to match the COM of the sacrum.All the listed changes resulted in a model used as “base” on which subject-specific characteristics of each dataset model were replicated thanks to a developed in-house Python language code. Further details on the changes apported to the model developed by Burkhart et al. and on the identified subject-specific characteristics are reported in the supplementary material.