Commercial buildings are responsible for a large fraction of energy consumption in developed countries, and therefore are targets of energy efficiency programs. Motivated by the large inherent thermal inertia of buildings, the power consumption can be flexibly scheduled without compromising occupant comfort. This temporal flexibility offers opportunities for energy savings and the provision of frequency regulation to support grid stability. To realize these goals, it is of prime importance to identify a realistic model for the temperature dynamics of a building. In this paper, we identify a low-dimensional data-driven model and a high-dimensional physics-based model for the same system at different spatial granularities and temporal seasons using experimental data collected from an entire floor of an office building on the University of California, Berkeley campus. We perform a quantitative comparison in terms of estimates of the inherent thermal gains due to occupancy, open-loop prediction accuracies, and closed-loop control schemes. We conclude that data-driven models could serve as a substitution for highly complex physics-based models with an insignificant loss of prediction accuracy for many applications.