LLM (Large Language Model)

LLMs like GPT, Claude, and Gemini are built by training a neural network — usually a Transformer — on trillions of words of text, code, and other data. The model doesn't “look up” answers; it learns statistical patterns of how language works well enough to generate coherent, often useful, text one token at a time.

Scale matters: bigger models trained on more data and more compute tend to perform better, which is part of why frontier labs spend billions of dollars on the GPUs and data centers needed to train them — spending that is now central to the debate over whether AI valuations have run ahead of AI revenue.