Hill models based on isometric force-length (FLR) and isotonic force-velocity relationship (FVR) fail to predict muscle force under dynamic conditions, due in part to absence of history-dependence of muscle force. To overcome this limitation, a new muscle model was developed with a titin element, based on titin-actin interactions in active muscle. The titin element wraps around a pulley, connecting to a contractile element in both series and parallel to simulate history-dependent forces. In this study, we aimed to test the ability of the new muscle model and a 12-parameter Hill model to predict forces observed during dynamic length changes. The strain of ex vivo mouse EDL was controlled using in vivo strain trajectories measured from guinea fowl LG during treadmill running over obstacles. Five strain trajectories (2 different upward strides, 1 downward stride, 1 level stride and 1 sine wave) and three activation patterns (Normal (in-vivo); Late (~12.5ms later start); and Long (duration + 33ms)) were used. The parameter set was trained using the ‘level stride’ strain trajectory and ‘Long’ activation and was tested on the other trials. The mean r² (0.73±0.12) was significantly lower for the Hill model (paired t-test, p < 0.01) than for the titin model (mean r² = 0.85±0.07). The result demonstrates that the new muscle model including a titin element predicts the dynamic variability of muscle force with higher accuracy than the Hill model based on FLR and FVR for strain trajectories typical of in vivo locomotion, and supports the idea that active titin-actin interactions contribute to muscle mechanics.