AI & Automation
Large Language Model
Definition
A large language model (LLM) is an AI system trained on vast amounts of text to understand and generate human language by predicting likely sequences of words.
What is a large language model?
A large language model (LLM) is a type of AI system trained on enormous quantities of text so that it can understand and generate human language. The "large" refers to scale in two senses: the volume of text it learns from, often much of the public web along with large collections of books and articles, and the number of internal parameters, frequently billions, that it uses to represent patterns in that text.
At its core, an LLM works by predicting the next word in a sequence. Given some text, it estimates what should come next, one piece at a time, and this simple mechanism, at sufficient scale, produces coherent explanations, summaries, translations, and answers. LLMs are the technology behind most current natural language processing tools.
How large language models are trained
Training usually happens in stages:
- Pre-training. The model learns general language patterns by predicting missing or next words across a vast body of text. This is where most of its broad knowledge comes from.
- Fine-tuning. The model is further trained on narrower, higher-quality data to make it better at particular tasks or to follow instructions.
- Alignment. Techniques such as learning from human feedback shape the model to be more helpful and to respond in expected ways.
Because a model's knowledge is fixed at training time, it does not automatically know about events, or about your business, after its training cut-off, which is one reason external sources are often supplied at the moment a question is asked.
What LLMs are good at, and their limits
LLMs excel at open-ended language tasks: answering questions, drafting and summarising text, translating, and explaining. Their great strength is flexibility, handling requests they were never explicitly programmed for.
Their central limitation follows from how they work. Because an LLM predicts plausible text rather than retrieving verified facts, it can produce a confident answer that is wrong, an AI hallucination. The practical fix is not to abandon the model but to ground it: supplying trusted, current sources at query time through retrieval-augmented generation, a form of AI grounding, keeps answers anchored to real content. In customer support, that is the difference between an assistant that sounds right and one that is right, because it answers from your documented knowledge rather than from memory alone.
Why it matters
Example
Ask a large language model to explain a refund policy and it will produce a fluent, well-structured answer in seconds. Given your actual policy document to work from, the answer reflects your real terms; left to rely on training alone, it may produce a reasonable-sounding policy that is not yours. The model's language is the same either way, which is why the source it draws on matters so much.
How Resolve247 helps
An LLM grounded in your knowledge base
Resolve247's AIChatbot puts a large language model to work on your support, but answers only from your knowledge base, with an anti-hallucination guarantee, so you get an LLM's fluency without the risk of invented answers.
30 day free trial, no cc required!
Related terms
Frequently asked questions
What is a large language model?
A large language model, or LLM, is an AI system trained on vast amounts of text to understand and generate human language. It works by predicting likely sequences of words, which lets it answer questions, summarise, translate, and explain.
How are large language models trained?
They are first pre-trained on a huge body of text to learn general language patterns, then fine-tuned on narrower data for particular tasks, and often aligned using human feedback. Their knowledge is fixed at the point of training, so up-to-date information is usually supplied when a question is asked.
What are large language models good at?
LLMs are strong at open-ended language tasks such as answering questions, drafting and summarising text, translating, and explaining concepts. Their main strength is flexibility, handling requests they were never explicitly programmed for.
How do you keep an LLM's answers accurate?
The most reliable approach is grounding: giving the model trusted, current sources to answer from, often through retrieval-augmented generation, rather than relying on training alone. Resolve247's AIChatbot uses this approach, answering only from your knowledge base.