What is a large language model (LLM)?
A large language model, or LLM, is an AI model trained on huge amounts of text to predict the next word in a sentence. Done at scale, that one skill lets it understand questions, summarise, translate and write fluent replies. GPT, Claude, Gemini and Llama are well-known examples.
Also called: LLM, foundation model, language model.
By Santhul Joseph, AI EngineerUpdated
How an LLM works
During training the model reads an enormous amount of text and adjusts billions of internal settings until it gets good at predicting what comes next. When you ask it something, it writes the answer piece by piece, each time picking a likely next word given everything before it.
It doesn't look facts up. It produces text that fits the patterns it learned, which is why it is so fluent, and also why it can be wrong.
What an LLM can't do on its own
- It doesn't know your business: your prices, opening hours or return policy.
- Its knowledge stops at the point its training ended.
- When it lacks a fact, it can still write a confident answer that is false. This is called a hallucination.
- It follows instructions in the text it reads, which people can try to abuse. This is called prompt injection.
How SimplyBoost uses an LLM
SimplyBoost doesn't rely on what the model remembers. It first finds the relevant passages in your knowledge, from website pages, PDFs, Word files, CSVs, images, Q&A pairs and more, and the model writes the reply from those.
Questions it couldn't answer are collected under Missing answers, so you can fill the gap, and the agent hands over to your team when it can't help.
Questions people ask
Can an LLM know about my business?
Not from its training. It knows your business when a system gives it your content at the moment it answers, which is what retrieval-augmented generation does.
Why do LLMs sometimes give wrong answers?
Because they predict likely text rather than check facts. Without a source to answer from, a likely-sounding sentence can still be false. Grounding the model in your own content makes this much less likely.
What does "large" mean in large language model?
It refers to the size of the model, often billions of internal settings called parameters, and to the huge amount of text it was trained on.