Precise flood forecasting is desirable so as to have more lead time for taking appropriate prevention measures as well as evacuation actions. Although conceptual prediction models have apparent advantages in assisting physical understandings of the hydrological process, the spatial and temporal variability of characteristics of watershed and the number of variables involved in the modeling of the physical processes render them difficult to be manipulated other than by specialists. In this study, two hybrid models, namely, based on genetic algorithm-based artificial neural network and adaptive-network-based fuzzy inference system algorithms, are employed for flood forecasting in a channel reach of the Yangtze River. The new contributions made by this paper are the application of these two algorithms on flood forecasting problems in real prototype cases and the comparison of their performances with a benchmarking linear regression model in this field. It is found that these hybrid algorithms with a “black-box” approach are worthy tools since they not only explore a new solution approach but also demonstrate good accuracy performance.