GPT

GPT (Generative Pre-trained Transformer) is a family — and now a general class — of large language models built on the decoder side of the transformer architecture. The name unpacks the recipe: generative (it produces text), pre-trained (it first learns broadly from a huge corpus before any task-specific tuning), and transformer (the attention-based network underneath). At its core it’s an autoregressive next-token predictor — given the text so far, it predicts the most likely next token, then repeats — which, at sufficient scale, yields fluent writing, reasoning, coding, and more. The lineage (OpenAI’s GPT-2/3/4 and successors) popularized the term, but “GPT” is now used loosely for decoder-only LLMs in general. Note the unrelated collision: GPT also stands for GUID Partition Table, a disk-partitioning scheme — context disambiguates.