A large language model (LLM) is an AI system trained on vast amounts of text to predict the next piece of text. By doing that extremely well, at enormous scale, it can answer questions, write, summarize, translate, reason through problems, and write code. LLMs power ChatGPT (OpenAI’s GPT models), Claude (Anthropic), Gemini (Google), and most other AI assistants.
Part of AI Basics. Last updated September 2026.
How LLMs Work in Four Steps
- Pre-training: the model reads a huge body of text and learns to predict the next word, absorbing grammar, facts, and patterns of reasoning.
- Fine-tuning: it is trained further to follow instructions and be helpful, often using human feedback.
- Your prompt: your message, plus any files or instructions, becomes the context the model responds to.
- Generation: it produces a response one piece at a time, each choice based on the context so far.
Key Concepts
| Term | Meaning | Why It Matters |
|---|---|---|
| Token | A chunk of text, roughly part of a word | Limits and pricing are measured in tokens |
| Context window | How much text the model can consider at once | Determines how long a document it can handle |
| Training cut-off | When the model’s training data ends | It may not know recent events without web search |
| Hallucination | A confident but incorrect answer | Always verify important facts |
| Reasoning models | Models that work through problems step by step before answering | Better for complex tasks, often slower |
What This Means for You
- Give context. The model only knows what is in its training and your prompt.
- Provide sources or use web search for current or specialist facts.
- Break big tasks into steps for better results.
- Verify anything important, especially numbers, quotes, and citations.
Frequently Asked Questions
Why do LLMs make things up?
They generate likely text rather than looking up facts, so when they lack information they can produce plausible but false answers.
Which LLM is best?
It depends on the task. See ChatGPT vs Claude and ChatGPT vs Gemini.
Do LLMs learn from my conversations?
Some consumer plans may use conversations to improve models unless you opt out; business plans typically do not. Check each provider’s settings.