A variety of software reliability growth models have been proposed to analyze the reliability of software, such as stochastic process models, neural network models and grey theory models. When dealing with the limited amount of failure data, researchers tend to seek for non-stochastic approaches because stochastic models require large samples of data to determine underlying distributions. However non-stochastic approaches fail to deal with more fluctuating data sequences. In this paper, we have used Grey-Markov Chain Model (G-MCM) and show the effectiveness of model in handling dynamic software reliability data. G-MCM combines the advantages of both grey prediction model (GM(1,1) and Markov Chain Model. GM (1,1) can robustly deal with incomplete and imperfect information and gives excellent prediction outcomes. At the same time, it takes the advantage of predictive power of Markov chain which decreases the random variability of the data and zpositively effects the forecasting accuracy. Musa’s failure datasets from various projects have been used to evaluate the prediction capability of G-MCM and compared with GM (1, 1) and Modified Jelinski-Moranda reliability prediction model. The comparison shows that the G-MCM has better prediction results than other models and has adequate applicability in software reliability prediction