Energy efficiency is a first-order objective for current Fifth Generation (5G) and emerging Sixth Generation (6G) radio access networks, where base stations dominate operational consumption and remain far from energy-proportional under time-varying load. In this context, the base-station power-consumption model is a first-order assumption: different formulations can yield materially different energy-saving estimates and, consequently, different energy-performance trade-offs. This thesis quantifies such model-driven effects through a controlled comparison between the standardized 3GPP TR~38.864 state-based model and an analytical Active Antenna Unit (AAU) formulation grounded in hardware-driven component scaling. A unified evaluation pipeline is designed and developed to ensure comparability: scenario, traffic, control knobs, and KPI definitions/normalization are kept fixed while only the power model is swapped.