Large Language Model (LLM)
A neural network trained on large text corpora that takes text input and produces text output. The underlying engine inside an AI agent, not the agent itself.
Also known as: language model · foundation model
What is Large Language Model (LLM)?
A large language model (LLM) is a neural network trained on large amounts of text that takes text input and produces text output. Leading examples include Anthropic’s Claude, OpenAI’s GPT, Google’s Gemini and Meta’s Llama. An LLM is the engine of an AI agent, not the agent: the agent is the prompt, knowledge base, tools and evals around it.
Christos Papadimitriou, theagency47 · Updated May 15, 2026LLM vs agent
This is the disambiguation that confuses buyers more than any other.
- The LLM is the model, a fixed artifact trained by Anthropic / OpenAI / Google. You access it via API. It does not know your business.
- The agent is the system built on top, system prompt + your data + tools + boundaries. It does know your business, because you configured it.
You do not “buy an LLM” for your business. You buy API access to one. Then an agency or your team builds an agent on top.
Which LLM theagency47 uses
We are model-agnostic and use a mix of LLMs, matched to each agent tier:
- Frontier reasoning models (Claude Opus, GPT, Gemini Pro) for Executive-tier agents (reasoning-heavy, low volume, high judgment)
- Balanced mid-tier models (Claude Sonnet, GPT, Gemini) for Operational-tier agents (balanced, production-scale, default choice)
- Fast, low-cost models (Claude Haiku, Gemini Flash, open-weight models such as Llama or Mistral) for Task-tier agents (fast, cheap per call, high volume)
Why a mix: no single provider leads on every job. We pick per task on tool-use reliability, language quality (including Greek), data-handling terms and cost, and route through a gateway so a model can be swapped without rebuilding the agent. Model choice is documented per agent and revisited yearly.
FAQ
Do I need to train an LLM on my data?
Almost never. Training (fine-tuning) is expensive and slow. Instead, you customize via system prompts and retrieval-augmented generation from a knowledge base. Same outcome, days vs months.
Why not just use ChatGPT directly?
ChatGPT is a consumer product. The Claude API is the production interface. Different access, different terms, different data handling.
Related terms
AI Agent
A software system that takes a goal, plans steps, calls tools, and acts autonomously to produce an output, without step-by-step human prompting.
System Prompt
The instructional text that defines an AI agent's role, voice, boundaries, and decision rules. The single most important customization layer of an agent.
Hallucination
An AI agent producing output that sounds confident but is factually wrong, typically because the model invented information not present in its inputs.
