The neural network architecture behind modern AI models, built on attention — letting every word directly consider every other word in parallel.
Introduced in 2017's "Attention Is All You Need," the transformer replaced reading text one word at a time with attention: each token looks at all the others at once and learns which matter. That parallelism fits GPUs perfectly, which let training scale to internet-sized data — and scaling is what made modern LLMs possible.
Nearly every model in this index — GPT, Claude, Gemini, Llama — is a transformer variant. Its known cost, attention's expense on very long inputs, drives the steady stream of "beyond transformers" research the index occasionally features.