Traditional Hill-type muscle models, parameterized using high-quality experimental data, are often “too weak” to reproduce the joint torques generated by healthy adults during rapid, high force tasks. This study investigated whether the failure of these models to account for different types of motor units contributes to this apparent weakness; if so, muscle-driven simulations may rely on excessively high muscle excitations to generate a given force. We ran a series of forward simulations that reproduced measured ankle mechanics during cycling at five cadences ranging from 60 to 140 RPM. We generated both “nominal” simulations, in which an abstract ankle model was actuated by a 1-element Hill-type plantar flexor with a single contractile element (CE), and “test” simulations, in which the same model was actuated by a 2-element plantar flexor with two CEs that accounted for the force-generating properties of slower and faster motor units. We varied the total excitation applied to the 2-element plantar flexor between 60 and 105% of the excitation from each nominal simulation, and we varied the amount distributed to each CE between 0 and 100% of the total. Within this test space, we identified the excitation level and distribution, at each cadence, that best reproduced the plantar flexor forces generated in the nominal simulations. Our comparisons revealed that the 2-element model required substantially less total excitation than the 1-element model to generate comparable forces, especially at higher cadences. For instance, at 140 RPM, the required excitation was reduced by 23%. These results suggest that a 2-element model, in which contractile properties are “tuned” to represent slower and faster motor units, can increase the apparent strength and perhaps improve the fidelity of simulations of tasks with varying mechanical demands.