An increase in the use of social media has led to a drastic increase in the amount of cyberbullying that takes place on such online platforms. The effects of online harassment can lead to both short-term and long-term mental health issues, depression, a lack of confidence, lower self-esteem and suicidal thoughts. In this paper, we suggest an effective approach for identifying such instances of cyberbullying on Twitter using transformers. Four different transformer architectures were trained and their results were compared with each other. BERT-Base-Uncased achieved a test accuracy of 85.81% and an F1-score of 0.8566, outperforming transformers such as DistilBERT-Base-Uncased, ELECTRA and MobileBERT-Uncased. Compared to traditional machine learning algorithms BERT-Base-Uncased produces better results, thus proving effective for use in the real-time identification of such malicious instances of cyberbullying.