How it writes
An LLM breaks text into small pieces called tokens. It predicts which piece is likely to come next, then repeats that step to form a reply. Training on large amounts of text helps it write, summarise, and answer questions.
An LLM is one part of an AI assistant. The app adds instructions, tools, and an interface around a model.
An app often sends your request to a model on the provider’s servers and shows you its reply. Some smaller models can run on your device. Where the model runs affects what data leaves your device and whether the feature can work offline.
Clear writing can still be wrong
An LLM can give a confident answer that is wrong or made up. It does not automatically check its statements against reliable sources. Important claims need checking, and products need stronger safeguards when mistakes could cause serious harm.