Future wireless communication systems being characterized by tight frequency reuse, adaptive modulation and coding schemes and diversified data services will be interference limited by the interfering signals of diverse rates and strengths. Frequency reuse factor of 1 being proposed for the next generation mobile systems will lead to 1 or 2 dominant interferers. We propose in this paper a low complexity maximum likelihood (ML) demodulator for interference suppression. The proposed demodulator is also applicable to general multi-stream (spatially multiplexed) MIMO systems where it reduces one complex dimension of the system i.e. the complexity of ML detection reduces from O (|chi|^eta) to O(|chi|^eta-1) where n is number of transmit antennas/spatial streams. This is a fundamental result as space and technology constraints shall be restricting future MIMO systems to low dimensionality. Therefore reduction of one complex dimension in detection shall enable implementation of ML detectors at the receivers. We look at the performance of linear MMSE demodulator and the proposed demodulator in the presence of interferers of diverse strengths and rates and simulation results demonstrate much improved performance of the proposed demodulator for interference suppression.