What is a large language model (LLM)?
A large language model (LLM) is an AI model trained on very large amounts of text to understand and generate human language. LLMs predict likely next words, which lets them answer questions, summarize, write, translate and reason, and they power chatbots and AI agents.
How LLMs work
Text is split into tokens. A neural network (usually a transformer) learns patterns from training data, then generates output one token at a time.
Strengths and limits
| Strengths | Limits |
|---|---|
| Fluent language | Can hallucinate |
| Broad knowledge | Knowledge cutoff |
| Reasoning | No company data unless given |
| Flexible tasks | Cost and latency |
In business
LLMs become useful when connected to company data (RAG, knowledge graphs) and tools (agents).
