Large Language Model

Noun · AI & Machine Learning · Origin: 2018

Definitions

  1. A neural network with billions of parameters trained on massive text datasets to understand and generate human language. GPT, Claude, Llama, and Gemini are LLMs. They don't 'think' in any meaningful sense — they predict the next token — but the results are uncanny enough to make that distinction feel academic.

    In plain English: A very large AI system trained on huge amounts of text that can understand questions and generate human-like responses — the technology behind ChatGPT and similar tools.

  2. In a systems architecture context, an LLM is a component that accepts text input and produces text output, often wrapped in an API. Engineers treat it as a probabilistic function with latency, cost, and reliability tradeoffs — not magic, just another service to integrate.

    Example: 'We swapped the LLM from GPT-4 to Claude in our pipeline and cut latency by 40% with no quality regression on our eval suite.'

    Source: engineering perspective

Etymology

2017
Google's 'Attention Is All You Need' paper introduces the Transformer architecture that makes LLMs possible
2018
OpenAI releases GPT-1 with 117M parameters; the term 'large language model' begins appearing in research papers
2020
GPT-3 at 175B parameters demonstrates that scale unlocks emergent abilities, cementing LLM as a category
2023
ChatGPT makes LLMs a household term, reaching 100 million users in two months

Related Terms