The current study attempts to investigate the performance of water based Mn-Zn ferrite magnetic nanofluid in a counter-flow double-pipe heat exchanger under quadrupole magnetic field using the two-phase Euler-Lagrange method. The nanofluid flows in the tube side as coolant, while the hot water flows in the annulus side. The effects of different parameters including concentration, size of the particles, magnitude of the magnetic field and Reynolds number are examined. Distribution of the particles is non-uniform at the cross section of the tube such that the concentration is higher at central regions of the tube. Application of the magnetic field makes the distribution of particles more uniform and this uniformity increases by increasing the distance from the tube inlet. Increasing each of the parameters of concentration, particle size and magnitude of the magnetic field will lead to a greater pressure drop and also higher heat transfer improvement. At higher Reynolds numbers, the effect of magnetic force is diminished. Optimization was performed using genetic algorithm coupled with compromise programming technique in order to reach the maximum overall heat transfer coefficient along with the minimum pressure drop. For this purpose, the models of objective functions of overall heat transfer coefficient and pressure drop of the nanofluid were first extracted in terms of the effective parameters using neural network. The neural network model predicts the output variables with a very good accuracy. The optimal values were obtained considering different conditions for relative importance of the objective functions. (C) 2013 Elsevier Ltd. All rights reserved.