Definition: large language model

The foundation of today's AI chatbots and agents. AI applications use a large language model (LLM) to understand natural language, answer questions and generate original content. An LLM is a type of artificial neural network (ANN) with billions of interconnections that analyze enormous quantities of text and images (see neural network).

The LLM is "trained" on data, sometimes massive amounts, and this learning phase takes the most computer time. Fine tuning the model makes it usable for different AI applications such as chatbots and text generation. See deep learning, AI weights and biases and AI transformer.

Part of a Robot
Along with electronic systems for physical motion, mechanical robots use language models for machine vision and natural language.

The World Is Online
The entire world's information is available over the Internet from websites, blogs and social media, from which trillions of data items have been extracted to train a large language model. See AI dataset.

Small, Edge and Reasoning Models
Designed for a specific purpose such as spam detection or website assistance, small language models (SLMs) have fewer layers and interconnections than LLMs. Edge language models are SLMs that run locally in a PC, and reasoning models tackle more complex problems than LLMs (see small language model, edge AI and large reasoning model).

All Over the Place

Today, language models with strange names and version numbers are popping up out of the woodwork. Sometimes overwhelming the AI enthusiast, the casual user has no clue why one is better than another without doing research or just trying them out. This partial list was presented in the Mammouth system. See Mammouth, GPT, Anthropic, Deepseek and Gemini chatbot.