A three-dimensional (3D) cellular automaton model is developed for the prediction of dendrite growth with stochastic orientation during solidification process in the molten pool of binary alloy. An angle-information transfer method is proposed for improving cellular automaton technique to simulate the growth of the dendrites whose preferred growth direction owns stochastic misorientation with respect to the direction of the coordinate system. Dendrite morphologies and solute distributions of single dendrite growth and multi-dendrite growth are able to be obtained by the simulation using present model. The model is also employed to study the difference between two-dimensional simulation and 3D simulation on solute segregation and dendritic growth. Using the established model, 3D multi-columnar dendrites with stochastic crystallographic orientations can be obtained efficiently, and the competitive growth and impinging of dendrites can be reproduced in practice. The simulation results agree well with the experimental results.