This paper investigates synchronization in a typical multi-agent system in which the communication network changes according to the system state. Through building new relationships between a matrix and its associated graph and estimating the diameter of the communication network, we prove that synchronization can be achieved if the speed of agents is bounded by O(n−β), where n is the number of agents and β is bounded by a constant independent of n, which is much better than the existing bound O(n−n). Some simulations are provided to illustrate the theoretical results.