FinBERT is a pre-trained language model built by further training BERT on a large financial corpus. It performs financial sentiment classification by fine-tuning the base BERT architecture for sequence classification tasks. The model is available on the Hugging Face Hub and relies on the transformers library for implementation.