This paper proposes an efficient list extension algorithm for soft-output multiple-input-multiple-output (soft-MIMO) detection. This algorithm extends the list of candidate vectors based on the vector selected by initial detection, in order to solve the empty-set problem, while reducing the number of additional vectors. The additional vectors are obtained from multiple detection orders, from which high-quality soft-output can be generated. Furthermore, a method to reduce the complexity of the determination of the multiple detection orders is described. From simulation results for a 4×4 system with 16- and 64-quadrature amplitude modulations (QAM) and rate 1/2 and 5/6 duo-binary convolutional turbo code (CTC), the soft-MIMO detection to which the proposed list extension was applied showed a performance degradation of less than 0.5dB at bit error rate (BER) of 10^-5, compared to that of the soft-output maximum-likelihood detection (soft-MLD) for all code rate and modulation pairs, while the complexity of the proposed list extension was approximately 38% and 17% of that of an existing algorithm with similar performance in a 4×4 system using 16- and 64-QAM, respectively.