Low power has become a primary concern in the field of ad‐hoc and personal area networks. As manufacturers start endowing their designs with scatternet support, Bluetooth is emerging as a key enabling technology. Although this is driving research on Bluetooth power optimization, most proposals are based on over‐simplified, fully theoretical, or old and inadequate power models. We present a real‐world power model of a Bluetooth device supporting scatternets and sniff‐mode, and validate it experimentally on a real implementation. Whilst guaranteeing a low computational complexity, the model achieves a 4% RMS error and can be a precious aid in the design and implementation of power‐critical Bluetooth applications.