A language model is a statistical model for sequences of words. When one such model has upwards of a few billion parameters, it is referred to as a "large language model", or LLM -- "large" here is vague in the same way as "big" in "big data".