The issue of sexism in online content has recently been a significant concern. With the increasing number of online interactions and the rise of social media platforms, the need for automated techniques to identify and classify sexism has become more critical than ever. This paper addresses this problem by fine-tuning deep-learning models for sexism classification using “MultiHate”. It is a comprehensive dataset created by curating ten different datasets on sexism. The dataset consists of 1.76 M English texts labelled as sexist and not sexist, then fine-tuned two deep learning models, Convolutional Neural Networks-Bidirectional Long Short-Term Memory and Generative Pre-trained Transformer 2, which accurately detect and classify sexism. A comparative analysis has been conducted on several machine learning and deep learning models using the MultiHate dataset. Investigation reveals that the Generative Pre-trained Transformer 2 model outperforms other models with an accuracy of 92%, while the Convolutional Neural Networks-Bidirectional Long Short-Term Memory model achieved an accuracy of 90% using precision, recall, and F1 scores as performance metrics. The models’ performances are promising, indicating that automated techniques can be employed to classify sexist content effectively. A comprehensive error analysis of the models’ performance has been presented, highlighting their limitations and challenges. The computational time required for training and testing the models is a significant challenge, especially for larger datasets.