We study the Boolean dynamics of the "quenched" Kauffman models with a directed scale-free network, comparing with that of the original directed random Kauffman networks and that of the directed exponential-fluctuation networks. We have numerically investigated the distributions of the state cycle lengths and its changes as the network size N and the average degree (k) of nodes increase. In the relatively small network (N similar to 150), the median, the mean value and the standard deviation grow exponentially with N in the directed scale-free and the directed exponential-fluctuation networks with (k) = 2, where the function forms of the distributions are given as an almost exponential. We have found that for the relatively large N similar to 10(3) the growth of the median of the distribution over the attractor lengths asymptotically changes from algebraic type to exponential one as the average degree (k) goes to (k) = 2. The result supports the existence of the transition at (k)(c) = 2 derived in the annealed model. (C) 2007 Elsevier Ltd. All rights reserved.