In this paper, we propose and compare three algorithms based on Bell Labs? vertical space-time codes (V-BLAST) and examine their block error rate and their algorithmic complexity for different combinations of block length and number of antennas. Of special interest are two algorithms that use least-squares techniques, specifically the QR decomposition, to substantially lower the complexity of the original V-BLAST. We present the algorithms in detail as well as simulation results and experimental results based on a high-performance digital signal processor. We show that the algorithm of choice depends on the number of antennas in the system, the block length L, and on whether the application at hand requires low-error rates or low-complexity.