What Is a Large Language Model (LLM)?

How LLMs Work

LLMs are built on the transformer architecture and contain billions of parameters. They are first pre-trained on broad text, then often fine-tuned and trained with human feedback to follow instructions and respond helpfully.

Limitations

Because they generate plausible text rather than looking facts up by default, LLMs can produce confident but incorrect statements, known as hallucinations. Their knowledge also reflects their training data up to a certain date.