This paper deals with the benefits of using a nonlinear model-based approach for controlling magnetically guided therapeutic microrobots in the cardiovascular system. Such robots used for minimally invasive interventions consist of a polymer binded aggregate of nanosized ferromagnetic particles functionalized by drug-conjugated micelles. The proposed modeling addresses wall effects (blood velocity in minor and major vessels' bifurcations, pulsatile blood flow and vessel walls, and effect of robot-to-vessel diameter ratio), wall interactions (contact, van der Waals, electrostatic, and steric forces), non-Newtonian behavior of blood, and different driving designs as well. Despite nonlinear and thorough, the resulting model can both be exploited to improve the targeting ability and be controlled in closed-loop using nonlinear control theory tools. In particular, we infer from the model an optimization of both the designs and the reference trajectory to minimize the control efforts. Efficiency and robustness to noise and model parameter's uncertainties are then illustrated through simulations results for a bead pulled robot of radius 250 μm in a small artery.