Which of the following statements best categorizes the advantages of the Transformer architecture compared to traditional RNNs in natural language processing tasks?

Transformers outperform RNNs mainly because they use self‑attention, which lets every word look at all other words at once, so the model can capture long‑range relationships quickly. Because attention does not rely on a sequential chain, the entire sentence can be processed in parallel, making training much faster on modern GPUs. This parallelism also reduces the risk of vanishing or exploding gradients that plague deep RNNs, allowing deeper networks and longer context windows. For example, in translating a sentence, a Transformer can instantly attend to the subject and verb across a long clause, while an RNN would have to propagate information step by step, often losing earlier context. Thus, Transformers combine speed, scalability, and stronger long‑distance context handling compared to traditional RNNs.