This paper proposes a novel recurrent neural network (RNN) model for hyperspectral image classification. The model uses a newly proposed activation function, parametric rectified tanh (PRetanh), and a modified gated recurrent unit (GRU) to effectively analyze hyperspectral pixels as sequential data. The PRetanh activation function allows for higher learning rates without the risk of divergence during training, while the modified GRU efficiently processes hyperspectral data and reduces the total number of parameters. Experimental results on three airborne hyperspectral images show competitive performance of the proposed model. The proposed network architecture opens new possibilities for future research in deep recurrent networks for hyperspectral data analysis. The model outperforms traditional vector-based methods such as SVM and random forest, as well as other RNN models like LSTM and GRU with tanh or ReLU activation functions. The proposed RNN achieves higher accuracy in classification tasks, particularly in distinguishing similar land covers. The model also demonstrates faster processing times compared to other methods. The results show that the proposed RNN is effective for hyperspectral image classification and has the potential to be a powerful tool for future research in this area.This paper proposes a novel recurrent neural network (RNN) model for hyperspectral image classification. The model uses a newly proposed activation function, parametric rectified tanh (PRetanh), and a modified gated recurrent unit (GRU) to effectively analyze hyperspectral pixels as sequential data. The PRetanh activation function allows for higher learning rates without the risk of divergence during training, while the modified GRU efficiently processes hyperspectral data and reduces the total number of parameters. Experimental results on three airborne hyperspectral images show competitive performance of the proposed model. The proposed network architecture opens new possibilities for future research in deep recurrent networks for hyperspectral data analysis. The model outperforms traditional vector-based methods such as SVM and random forest, as well as other RNN models like LSTM and GRU with tanh or ReLU activation functions. The proposed RNN achieves higher accuracy in classification tasks, particularly in distinguishing similar land covers. The model also demonstrates faster processing times compared to other methods. The results show that the proposed RNN is effective for hyperspectral image classification and has the potential to be a powerful tool for future research in this area.