Nowadays, people communicate through technology. Millions of comments are posted on social networks about a given topic. Users are becoming increasingly involved in spreading messages. Given the amount of unregulated content, it is so easy to post almost anything on social networks, which has given way to a proliferation of aggressive and harmful content commonly called hate speech. Hate speech on social networks has become a fundamental problem, affecting both community safety and online discourse. One of the most important challenges in Natural Language Processing (NLP) is detecting semantic analysis in texts. Conventional techniques for identifying such content frequently fail to capture the complex phrasing and context of hate speech. We propose designing and developing a model for the automated detection of hate speech on social networks. By utilizing the sophisticated powers of the Large Language Model (LLM) and NLP, this approach allows for more accurate comprehension of context and nuances than other models. We carried out a case study using a dataset of Panamanian users extracted and labelled by us to evaluate the model, achieving excellent results. The outcomes demonstrate a significant improvement, with our model outperforming conventional classifiers by achieving over 82% accuracy. This development gives social network sites a considerable advantage in reducing hate speech and is in line with current advances in using deep learning for complex linguistic tasks. Analysis of the model’s errors reveals a number of miscategorised tweets, potentially due to underfitting caused by the limited training data. However, as the amount of training data increases, the number of these errors decreases.