LLM is the underlying “horsepower/core” of today’s AI products. Today’s popular LLMs include GPT 5.5, Claude Opus 4.7, Gemini 3.1, etc. (See LLM Model Names for more details, as the naming can be confusing).
There are two separate layers: the LLM and the product/use case, not to be confused.
Analogy: LLM are like engines. The same engine can be used for e.g. sports car, truck, motorcycle, like how the same LLM can be used in different apps like ChatGPT, Copilot, Cursor. We can also swap different engines in the same car, like how we can pick different LLMs within the same app e.g. Copilot.
Practical example: Notion AI. The use case here is an agent that helps you organize your Notion docs. We can pick the underlying LLM, but they will all serve the same role.
Different LLMs excel at different areas. Some LLMs are good at coding, but perhaps less capable at writing. It is important to pick the suitable LLM for the use case. E.g. for coding tasks, pick LLM thats good at coding! See Choosing an LLM for details.
Ignore for now
You don't need to know how an LLM works internally to use them. Example terms include transformer, attention.
If you're interested in digging deeper, here are a few intuitive explanations: