A large language model learns statistical patterns across large text and code datasets, then predicts tokens to produce responses. Modern models can also use tools, process additional modalities, and participate in multi-step agent workflows.
Model capability depends on architecture, training, context, inference software, and hardware. A strong model can still run poorly when kernels or serving systems do not match its operations efficiently.
ELI5
A large language model is an AI model trained on a very large amount of tokenized text and often code. It learns patterns that let it continue language, answer questions, create text, and produce other structured results.
For example, a model can read a question and generate a useful explanation one token at a time. Its answer can still be wrong because predicting likely language is not the same as checking every claim against reliable evidence.
