For the last few decades, many papers focused on rare but significant large-loss events. There are many kinds of models for portfolios. In this paper, we establish an importance sampling method for t-copula model. In t-copula model, because of the heavy tail of the t-distribution, we can not handle the moment generating function which is useful in exponential tilting. To solve this problem, we make 3-step importance sampling method. First step is importance sampling for a chi-squared random variable and use 2-step importance sampling method which established by Glasserman, Kang, and Shahabuddin [4].