Generative pre-trained transformer
OpenAI's name for its series of decoder models, which has become a general label for models of this kind.
Also: GPT
GPT stands for generative pre-trained transformer: a transformer, pre-trained on unlabelled text, that generates text. OpenAI introduced the approach with GPT-1 in June 2018, followed by GPT-2 in 2019, GPT-3 in 2020 and GPT-4 in 2023, and the letters appear in ChatGPT. The term is widely used for similar models from other makers, and many open models include it in their names. OpenAI tried to register GPT as a trademark in the United States but was refused in 2024 on the grounds that it was too descriptive.