A large language model is an AI system trained on vast amounts of text data to understand and generate human language. Models like GPT-4, Claude, and Gemini are examples. They are called "large" because of the scale of both the training data and the number of parameters, the internal numerical weights that define how the model processes and generates text.
LLMs work by learning statistical relationships between words, phrases, and concepts across billions of examples. Given a prompt, the model predicts the most contextually appropriate response, one token at a time. The result is a system that can write, summarize, translate, answer questions, classify content, generate code, and engage in multi-turn conversation with a level of fluency that was not achievable with earlier natural language processing approaches.
For organizations, LLMs are most valuable when combined with specific data, tools, and guardrails that focus them on a defined task. A general-purpose LLM becomes an enterprise-grade system when paired with a retrieval layer (RAG), clear system instructions, integration with company data sources, and governance controls that define what the model is and is not permitted to do.
The major hosted LLMs are available via API from Anthropic, OpenAI, Google, and others. Organizations choose between them based on capability benchmarks, pricing, context window size, latency, safety controls, and data handling agreements.