The internet has emerged as a central platform for public discourse and opinion sharing, with social media significantly enhancing human interaction. To counteract the detrimental effects of online hate speech, precise detection and evaluation are essential. This paper provides a thorough assessment of hate speech detection in Brazilian Portuguese, utilizing BERT and traditional machine learning models as benchmarks. Although BERT demonstrated superior performance in hate speech classification, our results also indicate that LSTM-based models can effectively contribute to this task, presenting a viable alternative.