This letter presents novel Expectation Propagation (EP) algorithms for Single User (SU) and Multiple Users (MU) Multiple-Input Multiple-Output (MIMO) detection that address the performance loss and slow convergence of classic Scalar EP (SEP) solutions with even lower computational complexity. The proposed algorithms employ QR Decomposition (QRD), optimized EP message scheduling, and Successive Interference Cancellation (SIC) techniques to achieve superior performance and convergence speed comparable to those of Vector Expectation Propagation (VEP). Our results are illustrated through error rate curves and analysis of performance-complexity trade-offs in SU and MU scenarios.