One of the essential phases in water resource planning and management is streamflow forecast. It is necessary for the functioning of hydropower plants, agricultural planning, and flood control. The present study applied Artificial Neural Network (ANN) model, Adaptive Neuro-Fuzzy Inference System (ANFIS), Bidirectional LSTM (BiLSTM), and hybrid Convolutional Neural Network (CNN) Gated Recurrent Unit (GRU) Long-Short Term Memory (LSTM) model to predict the long-term daily streamflow in the Colorado River, USA. 60% of the data (1921–1981) was used for training, while 40% of the data (1981–2021) was utilized for testing the model's performance. The obtained outcomes of the suggested models were assessed using four assessment indices, including by Normalized Root Mean Square Error (NRMSE), Mean Absolute Error (MAE), Correlation Coefficient (r), and Nash–Sutcliffe Coefficient (ENS). Based on the comparison of outputs, in the testing phase, it was determined that the ANFIS model with
NRMSE = 0.116, MAE = 24.66, r = 0.968, and ENS = 0.936 outperformed the other studied models in terms of reliability and accuracy. While the CNN-GRU-LSTM and BiLSTM models are complex, they do not perform better. The comparison demonstrates that the performance of their respective models is not much better than the two standard models-ANN and ANFIS.