Models based on attention, such as transformers [36], changed how we process sequences. Unlike older recurrent neural networks that handle sequences step by step, transformers use attention to examine the entire sequence at once and do so much faster. Recurrent models often struggle with long-term dependencies because they process tokens in sequence. Transformers avoid this limit. They model relations between any two tokens, no matter how far apart they are. This design improves efficiency and accuracy in many NLP and sequence tasks. Today, transformers are a core tool in deep learning.