The use of long-term mechanical circulatory support (MCS) for heart failure by means of implanted continuous-flow left ventricular assist devices (cf-LVADs) will increase, either to enable recovery or to provide a destination therapy. The effectiveness and user-friendliness of MCS will depend on the development of near-physiologic control strategies for which accurate estimation of pump flow is essential. To provide means for the assessment of pump flow, this study presents pump models, estimating pump flow (Qlvad) from pump speed (n) and pressure difference across the LVAD ([DELTA]plvad) or power uptake (P). The models are evaluated for the axial-flow LVADs HeartAssist5 (HA5) and HeartMate II (HMII), and for a centrifugal pump, the HeartWare (HW). For all three pumps, models estimating Qlvad from [DELTA]plvad only is capable of describing pump behavior under static conditions. For the axial pumps, flow estimation from power uptake alone was not accurate. When assuming an increase in pump flow with increasing power uptake, low pump flows are overestimated in these pumps. Only for the HW, pump flow increased linearly with power uptake, resulting in a power-based pump model that estimates static pump flow accurately. The addition of pressure head measurements improved accuracy in the axial cf-LVAD estimation models.