Focusing on high order multiple-input multiple-output (MIMO) broadcast systems, this paper presents an efficient group competition-based user selection (GCUS) algorithm for multiuser beamforming. Motivated by the fact that the best user who has the strongest channel gain can not be guaranteed to be selected as one member of the optimal user set, in this paper, we treat the good users who have the relatively large channel gain as the strong candidates for the optimal user set. To pick out these candidates, firstly a gain window is created, which is controlled by a threshold. Secondly, several group leaders can be selected by the gain window and a group set is obtained as well. Then, based on its group leader, each group selects its group members independently using a Gram-Schmidt orthogonalization-based algorithm. Finally, the best group having the biggest sum rate is selected as the final user set. Besides multiuser diversity gain, the GCUS offers grouping gain that is obtained by selecting one group in a group set. The sum rate performance evaluation of the proposed algorithm is given by computer simulation and its complexity is evaluated by measuring the total number of floating-point operations (flops).