Skip to main content
All terms
Models & Products

GPT-3

OpenAI's 2020 model that showed a single model could do tasks from just a few examples in the prompt.

Definition

GPT-3 is a large language model released by OpenAI in 2020 with 175 billion parameters (the internal numbers it learns during training). It showed that a single pretrained model could handle many language tasks just from a prompt — given a few examples (few-shot) or none at all (zero-shot) — without separate training for each task. Accessed via OpenAI's API, it underpinned early applications including the Codex code model, and helped launch the current era of large language model development.