Large Language Model (LLM)

A large language model (LLM) is a neural network trained on massive text datasets to understand and generate human language. These models contain billions of parameters -- numerical weights learned during training -- that encode patterns of grammar, facts, reasoning strategies, and even coding conventions.

How It Works

LLMs are built on the transformer architecture, which uses a mechanism called self-attention to weigh the relevance of every word (or token) against every other word in the input. During training, the model learns to predict the next token in a sequence, and through this deceptively simple objective, it develops broad capabilities in text comprehension, translation, summarization, code generation, and logical reasoning.

When you send a prompt to an LLM, the model processes your input tokens through dozens of transformer layers, then generates a response one token at a time. Each new token is selected based on probability distributions shaped by the model's parameters and settings like temperature.

Why It Matters

LLMs are the foundation of modern AI assistants and agents. Every AI-powered coding tool, chatbot, and automated workflow depends on an LLM to interpret instructions and produce coherent outputs. The choice of LLM -- its size, training data, and capabilities -- directly affects the quality, speed, and cost of your AI deployment.

In Practice

When deploying an AI assistant, you select an LLM through a model provider such as OpenAI, Anthropic, Google, or an open-source alternative. The model is specified in your configuration, and the deployment platform routes requests to the appropriate provider's API. Understanding your LLM's strengths, context window limits, and pricing model is essential for building reliable AI-powered applications.