Automatic translation of indigenous languages poses a significant challenge due to the scarcity of digital resources and extensive parallel corpora. This study presents a Transformer-based model for the automatic translation of Náhuatl to Spanish, aiming to contribute to the preservation and accessibility of this language. A parallel corpus of Nahuátl-Spanish texts was used to train the model, optimizing its parameters to enhance translation quality. The evaluation was conducted using BLEU and ROUGE metrics, comparing the results with traditional approaches. The results indicate that, although the model successfully captures basic syntactic structures, it struggles with lexical ambiguity and the complex morphology of Náhuatl. This work demonstrates the potential of NLP models for processing low-resource languages and lays the foundation for future improvements in the automatic translation of indigenous languages.