The emerging service grids bring together various distributed services to a ‘market’ for clients to request and enable the integration of services across distributed, heterogeneous, and dynamic virtual organizations. In the experience of constructing and using the ChinaGrid, we meet two challenges, optimizing the QoS of the grid resources and minimizing complexity for application users and developers. In this paper, we present Q-SAC, a QoS optimized service automatic composition model to address these problems. Two main features of Q-SAC are (1) automatic grid service composition, and (2) global level multidimensional QoS optimization for the composition plan. We design the algorithms for generating and optimizing the composite services. The simulation results show that our model and solution are practical and efficient.