What Is an LLM? Large Language Models in Plain English

What does LLM mean?

LLM means large language model. 'Large' refers to the size of the model (often billions of parameters) and of the training data; 'language model' means it models how text is written so it can predict what comes next.

Source: Wikipedia: Large language model

Is ChatGPT an LLM?

ChatGPT is an app built on LLMs. The models underneath, such as OpenAI's GPT-6 family, are the LLMs; ChatGPT adds the chat interface, memory, tools like web search and image generation, and safety systems. The same is true of Claude, Gemini and Meta AI.

Source: OpenAI: GPT-6 Astra, a new generation of intelligence

What is the difference between AI, generative AI and an LLM?

AI is the broad field. Generative AI is AI that creates new content such as text, images, audio or video. An LLM is one kind of generative AI that works with language. Image generators and video models are generative AI but not LLMs, although many modern models handle several of these at once.

Source: Wikipedia: Large language model

How is an LLM trained?

First, pre-training: the model reads a huge amount of text and repeatedly adjusts its parameters to get better at predicting the next token. Then fine-tuning and reinforcement learning from human and AI feedback teach it to follow instructions, hold a conversation, refuse harmful requests and give more useful answers.

Source: Wikipedia: Large language model

What are tokens and context windows?

LLMs read and write text in tokens, small chunks that are often parts of words. The context window is the maximum number of tokens the model can consider at once, including your prompt, any documents, the conversation so far and its reply. Usage-based pricing is usually per million tokens.

Source: Wikipedia: Large language model

Do LLMs understand what they write?

They learn rich statistical patterns that let them reason through many problems, but they do not check facts against the world unless they are given tools such as web search or your documents. That is why grounding answers in sources matters.

Source: Wikipedia: Large language model

What are the main LLMs today?

Frontier closed models include OpenAI's GPT-6 family (such as GPT-6 Astra), Anthropic's Claude models, Google's Gemini 4 and Meta's Muse Spark. Popular open-weight models you can download and run include OpenAI's gpt-oss, Google's Gemma 4, Qwen, NVIDIA Nemotron and Meta's Muse Glimmer.

Source: OpenAI: GPT-6 Astra, a new generation of intelligence; Google: Gemini 4 Argon, our next era of frontier intelligence (Sep 30, 2026); Ollama Blog

What is a local LLM?

A local LLM is an open-weight model running on your own computer with a tool such as Ollama or LM Studio. It keeps data on your device and has no per-message cost, at the price of needing enough memory and usually being less capable than the largest cloud models.

Source: Ollama: run open models locally; LM Studio

What is an LLM hallucination?

A hallucination is when the model states something false, such as a made-up fact, quote, citation or statistic, in a confident tone. It happens because the model generates plausible text rather than looking facts up. Ask for sources, use tools that ground answers in documents, and verify anything important.

Source: Wikipedia: Large language model

Are LLMs biased?

They can reflect biases in their training data and in how they were tuned. Providers work to reduce this, but it is worth checking outputs on sensitive topics such as hiring, health, legal and financial decisions, and keeping a human in the loop.

Source: Wikipedia: Large language model

How do I write good prompts for an LLM?

Give the model a role, the task, the context it needs, the format you want and any constraints, plus an example if you can. Our prompt engineering guides and free prompt library show tested patterns for writing, research, coding and business work.

Which LLM should I use?

For general writing and research, a frontier assistant such as ChatGPT, Claude or Gemini is easiest. For coding, compare the leading coding agents. For privacy or offline use, run an open-weight model locally. Compare costs on our LLM pricing page before committing to API usage.