Large Language Model (LLM)
A type of AI model trained on large volumes of text that can understand and generate language. LLMs are the foundation of AI agents, copilots, and document processing tools — they are what reads the invoice, understands the supplier email, and drafts the exception summary.
What is a Large Language Model (LLM)?
A large language model is an AI system trained on billions of words of text to understand and generate human language. The "large" refers to both the size of the training data and the number of parameters — the internal weights that determine how the model responds to any given input. LLMs are the underlying technology behind AI agents, AI assistants, document processing systems, and most enterprise AI tools deployed since 2022.
For operations teams, an LLM is the component that reads unstructured text — a supplier email, a PDF invoice, a delivery note in Dutch — and turns it into structured information an ERP or workflow system can act on. It is also what drafts the exception summary, the supplier message, and the approval recommendation.
Context Window: What the Model Can See at Once
One of the most operationally significant differences between LLMs is the context window — how much text the model can process in a single interaction. Larger context windows mean the model can work with longer documents, more history, and more supporting data at once without losing earlier information.
Model | Context Window | Operational Note |
|---|---|---|
GPT-4o (OpenAI) | 128K tokens (~90K words) | Strong general-purpose; broad tool support; widely integrated in enterprise platforms |
Claude 3.7 Sonnet (Anthropic) | 200K tokens (~150K words) | Strong on complex reasoning and long documents; native MCP support; good for multi-step agent tasks |
Gemini 1.5 Pro (Google) | 1M tokens (~700K words) | Largest context available; suited for processing entire contract repositories or document archives in one pass |
Llama 3 (Meta, open-source) | 8K–128K tokens (varies) | Deployable on-premise; relevant for operations with strict data residency requirements |
LLMs in Operations
In practice, you rarely interact with an LLM directly. It sits behind the AI agent, the document processor, or the workflow tool you are using. But understanding the model underneath matters for three reasons:
Capability ceiling: The model determines what the system can understand and generate. A weak base model produces worse outputs regardless of how well the surrounding system is designed.
Context limits: If your AI agent needs to process a 200-page procurement framework contract, the model's context window determines whether it can do that in one pass or needs to be chunked — which affects accuracy.
Data handling: Different models have different data residency and privacy terms. For operations handling sensitive supplier contracts or financial data, this is an infrastructure decision, not just a technology one.