In recent years, the application of deep learning models has significantly advanced the field of computer vision, enabling automated recognition and classification of various objects, including flowers. This research begins with exploring two distinct pre-trained convolutional neural networks (CNNs): VGG16 and Xception. Each model has architecture and performance characteristics that are analyzed and compared to establish a baseline for flower species classification. To enhance classification performance further, we introduce a hybrid model that fuses the extracted features from VGG16 and Xception. These features are concatenated and fed into a dense layer with ReLU activation, followed by a softmax classifier, which leverages the combined knowledge of hybrid models to classify various species of flowers accurately. Experimental results are presented on a benchmark flower dataset from Kaggle, demonstrating the effectiveness of the proposed hybrid model in achieving state-of-the-art classification accuracy. The results highlight the performance of the proposed hybrid model for 25 epochs with 512 dense layers, showcasing a remarkable state-of-the-art classification accuracy of 91.20% on the Kaggle flower dataset. The comprehensive evaluation includes quantitative metrics such as accuracy, precision, recall, and F1-score, highlighting how robust the model is and its generalization capabilities. The findings in this research can assist in developing deep learning-based flower species classification systems.