Numerical modeling is widely used in structural engineering to represent buildings response under seismic loading conditions. However, even though numerical modeling is a common tool to characterize the behavior of structures, modeling uncertainties can lead to a broad range of expected response, particularly when representing the behavior of novel systems or components. Addressing different modeling choices can provide more informed insights into the response of structures, especially prior to conducting experimental tests or participating in blind prediction contests. Herein, blind response prediction of a novel steel system was conducted before testing at the E-Defense facility in Japan. The full-scale specimen consisted of a weak Moment-Resisting Frame (MRF) retrofitted with steel spines and force-limiting connections (FLC). The set of pre-test predictions involved addressing of different modeling choices to overcome the many sources of epistemic uncertainties and to provide greater confidence in the design and experimental testing program. Several models were subjected to the records specific to the testing program (Northridge Sepulveda and JMA Kobe) to estimate drift and acceleration responses. Numerical results were compared to the experimental data from the shake-table tests. Although all the models were able to represent general trends in drifts and accelerations and enabled proper development of the testing plan, peak response varied significantly depending on the modeling choices, especially those altering the system’s natural periods or those leading to different yielding patterns.