The transformer is the neural network architecture introduced by Google Research in the 2017 paper "Attention Is All You Need." It replaced recurrent networks (RNN, LSTM) as the dominant architecture for language modelling and now underlies every major large language model — GPT, Claude, Gemini, Llama, Mistral.
The core innovation is the self-attention mechanism, which lets the model weigh the relevance of every input token against every other token in parallel. This unlocked training on massive datasets across thousands of GPUs, which unlocked scaling laws, which unlocked the modern LLM era.
Transformers now power text, image, audio, and multi-modal models — the architecture is fundamental to nearly all 2020s AI progress.