Soldiers undergo regular physical testing to assess their functional capacity. However, current physical tests, such as push-ups, sit-ups, and pull-ups, do not necessarily assess job-specific physical capability. This article assesses the utility of generic predictive tests and a task-related predictive test in predicting performance against four job-critical military manual handling tasks. The box lift and place test was found to be the superior predictor in performance of four job tasks; a pack lift and place (R2 = 0.76), artillery gunner loading simulation (R2 = 0.36), bombing up an M1 tank simulation, (R2 = 0.47) and a bridge building simulation (R2 = 0.63). Pull-ups and push-ups were poor predictors of performance in the majority of job tasks. Although the box lift and place had a larger correlation with the artillery gunner loading task than the generic assessment, it only accounted for 36% of the variance, indicating that a task simulation may be more appropriate to assess soldiers’ capacity to perform this job task. These results support the use of a box lift and place rather than generic fitness tests for the evaluation of military manual handling tasks.