In a globalized world, machine translation between languages, especially between Spanish and Quechua, is crucial for promoting cooperation and preserving endangered languages. The lack of effective translation tools limits Quechua speakers’ access to modern resources, potentially leading to their marginalization. Neural network technology has improved machine translation, captured complex nuances, and facilitated the preservation of indigenous languages. This research develops a natural language processing (NLP) model for Spanish-Quechua translation using the “somosnlp-hackathon-2022/spanish-to-quechua” corpus. Two models, opus-mt-es-en and t5-small, were evaluated using various hyperparameter configurations. The results indicate that the opus-mt-es-en model achieved a maximum BLEU score of 15.47% and a CHRF score of 49.80%. In comparison, the t5-small model showed inferior performance, with a maximum BLEU score of 2.81% and a CHRF score of 20.82%. Finally, it is concluded that the implementation of this NLP model will improve the communication and preservation of Quechua by integrating it into digital platforms and promoting cultural diversity. Opus-mt-es-en was more adequate than t5-small, highlighting the importance of correctly adjusting the models and their hyperparameters for effective translation.