This research aims to address the challenge of classifying a diverse range of 10 commercial starchy corn varieties in Cusco (Peru). This is a significant challenge due to the shared morphological characteristics and hybridization tendencies among different varieties and ecotypes. To tackle this issue, an automatic classification hybrid model was developed that combines the strengths of a MobileNetV2 Convolutional Neural Network (CNN) for efficient feature extraction with a Support Vector Machine model for precise classification. The hybrid model is noted for its resource efficiency, lightweight architecture, hierarchical feature detection capabilities, and adaptability to the different morphological aspects of corn varieties. Furthermore, it demonstrates increased robustness in classification and reduction of overfitting, leading to an efficient performance with a validation accuracy of 99.2% in variety identification. This represents a significant improvement over traditional classification methods. The efficiency achieved in variety identification suggests its practical application in agronomy, it could be adapted to optimize other processes such as agricultural forecasting. Additionally, the application of this hybrid model could be extended to future research in the automatic classification of other Andean cereals, such as quinoa and kiwicha.