Neuronal gamma-band synchronization is a common phenomenon found in cortical networks, which is considered as a potential mechanism for communication among brain areas. How neural assemblies transit information within the narrow time window of each gamma cycle is still an open question. Previous modeling studies have demonstrated that precise spike timing can robustly carry information with the propagation of strongly synchronized spikes. Here we show that the temporal structure of loosely synchronized spikes within each gamma cycle can also effectively carry information in the noisy cortical networks. The relative spiking phase of the synchronous spikes are significantly more consistent under the same stimulus compared to those in random stimuli. Moreover, there is an optimal conduction delay distribution for the network to maximize the information transmission. Our work suggests that the loosely synchronized spikes in the gamma cycles may provide a fundamental mechanism for neural communication using temporal codes.