Transformer

A specific type of neural network architecture introduced in 2017, using a self-attention mechanism that allows processing of an entire sequence of tokens simultaneously. Used in state-of-the-art LLMs like GPT.