This paper focuses on value-at Risk (VaR) models for Taiwan call warrant writers. We consider Vega risk and add it into Delta-Normal Method and Monte Carlo simulation Method. We try to find an appropriate VaR model for the call warrant writers in Taiwan. the results of backtesting show that the VaR model considering Vega risk demonstrates better performance, no mater it is Delta-Normal Method or Monte Carlo simulation Method. Besides, we also find that for the deep-out-of-money call warrants, Monte carlo simulation Method could be better. However, as to the at-the-money call warrants, Delta-Normal Method is a superior VaR model.