Definition
Large language model is a large language model (LLM) is an AI system trained on huge amounts of text to predict the next token, which lets it write, summarize, translate, answer questions and generate code.
A large language model, or LLM, is the kind of AI behind chatbots such as ChatGPT, Claude and Gemini. It is a program trained on a very large amount of text and code to do one thing well: predict what comes next. That one skill, repeated millions of times, turns out to be enough to write an email, explain a concept or draft a function.
Text is first split into tokens, small chunks that are often a word or part of one. The model reads the tokens you give it and produces a probability for every possible next token. It picks one, adds it to the text, and repeats until the answer is done. Settings such as temperature control how adventurous the picks are.
What the model knows is stored in its parameters, also called model weights: billions of learned numbers set during pretraining. Most modern models use a design called the transformer. After pretraining, post-training steps such as fine-tuning and learning from feedback teach it to follow instructions and behave like an assistant.
Large refers to the size of the model and the amount of data and computing power used to train it. The word is relative. A model that counted as large a few years ago can be small today. Size usually helps, but it is not the only factor: training data quality and post-training matter a great deal.
An LLM is strong at language tasks: drafting, rewriting, summarizing, translating, classifying and coding. It is weaker at anything that needs guaranteed facts or exact arithmetic, because it generates plausible text rather than looking facts up. This produces hallucinations, fluent answers that are false. Giving the model your documents through retrieval-augmented generation, or letting it call a calculator or search through tool use, reduces the problem.
It also has a knowledge cutoff, a date after which it knows nothing unless you provide the information, and a context window that limits how much text it can consider at once.
Almost every AI writing, coding and chat product you use runs on an LLM. Knowing what it is helps you judge its answers: treat the output as a strong draft that needs checking, not as a database.
Newer systems add images and audio, which makes them multimodal models. Others are trained to think before answering, which makes them reasoning models. When an LLM is given goals and tools and acts over many steps, it becomes an AI agent.
An LLM is the underlying model that predicts text. A chatbot is an app built around an LLM, adding an interface, instructions and sometimes tools.
They model patterns in language well enough to produce useful text, but they generate likely continuations rather than looking up verified facts, so they can state false things confidently.
Mistral released Voxtral 3 on Oct. 20: new transcription and speech models plus a voice agent API at $0.04 a minute when it reasons with Mistral Small 4. Skip the benchmark charts. The per-minute price and the choice of where audio is processed are the story.
Mistral opened a public preview of Mistral Large 4 on Tuesday, a mixture-of-experts model with about 1 trillion parameters and 49 billion active, and promised weights by the end of October. The API is a preview. The weights are the launch.
NorForge found more than 20 exposed panels for the Cairn AI hacking tool, one cluster running over 1,100 attack projects at an estimated $5 a task. A 10% hit rate at that price means attackers can afford to try every store online.