Despite recent advancements in predicting upper-limb exoskeleton support for static postures, understanding the torque-angle relationship during dynamic movements remains challenging due to the lack of dynamic movement data and potential hysteresis effects. This study quantified the supportive torque provided by passive upper-limb exoskeletons during dynamic movements, specifically shoulder flexion versus extension, using the Ottobock Paexo Shoulder V2 and Ekso EVO. The study used the HUMAC NORM Isokinetic Dynamometer to measure torque at 15°/s, and polynomial regression models were developed to predict torque for all device support settings (mean adjusted R2 of 0.99). Results showed the Ottobock provided 39 ± 9% more support during shoulder extension, and the Ekso provided 42 ± 11% more support, with the Ekso displaying a wider range of peak torques (2.85–18.58 Nm) compared to the Ottobock (3.68–9.65 Nm). Both exoskeletons exhibited reduced torque output during dynamic shoulder flexion compared to extension due to hysteresis effects. To mitigate side effects from these torque differences, the exoskeleton support level should be set lower than the user’s upper-limb weight, using the dynamic extension equations. Overall, these findings offer valuable insights for future dynamic exoskeleton analyses. The provided equations can be utilized in musculoskeletal modeling and digital human modeling software to examine the effects of exoskeleton usage during work tasks involving dynamic upper-limb movements.