Local LLM
Short answer
A local LLM is a language model whose weights sit on your own disk and whose answers are computed by your own processor, with no request to anyone's server.
The trade is capability for control. A local model that fits on a consumer GPU has single-digit billions of parameters against the frontier models' hundreds of billions, so it loses badly on hard reasoning and obscure knowledge.
On the task of reading a document that is already in front of it, the gap narrows sharply, because most of the work is retrieval and comprehension rather than recall. That is why a browser is a good place for a small model and a research assistant is not.
Related terms
Other ways people ask this
- what is local llm