# Large language model (LLM)

> A large language model (LLM) is an AI system trained on massive amounts of text data that can understand, reason about, and generate human-like language. LLMs power a wide range of applications, from search assistants to content tools, by predicting and producing contextually relevant text.

_Source: https://nobi.ai/glossary/large-language-model_

## What is Large language model (LLM)?

LLMs are neural networks with billions of learned parameters that encode patterns from books, websites, and other text sources. They can answer questions, summarize content, translate languages, and hold multi-turn conversations. Because they learn from general text, they work across many domains without task-specific programming. Most commercial AI assistants and search tools today are built on top of one or more LLMs.

## How does large language model work?

- The model is pre-trained on a large corpus of text, learning statistical relationships between words and ideas.
- It is then fine-tuned or prompted to behave in a specific way - for example, as a customer assistant or a coding helper.
- When a user submits a query, the model processes the input and generates a response word by word, drawing on what it learned during training.
- Many production systems ground the model in a specific knowledge base so it answers from verified content rather than relying on training data alone.

## Why does it matter?

For ecommerce operators and dealerships, LLMs make it possible to handle a large volume of natural-language shopper or buyer questions without scaling a human support team. They can surface product details, explain inventory, and guide decisions in the same conversational way a knowledgeable sales associate would. Choosing a solution built on a well-grounded LLM reduces the risk of the system inventing answers that do not match your actual catalog or policies.

[Nobi](https://dashboard.nobi.ai) uses language models to understand shopper and buyer queries in natural language and generate answers grounded in the content and data you connect - so responses reflect your actual inventory, policies, and products rather than generic training data.

## Frequently asked questions

**What is the difference between an LLM and a traditional search engine?**
A traditional search engine matches keywords and returns a ranked list of links. An LLM interprets the intent behind a question and generates a direct, conversational answer - it understands context and follow-up questions rather than just pattern-matching on words.

**Can an LLM give wrong answers?**
Yes. LLMs can produce plausible-sounding but incorrect information, especially on topics not well represented in their training data. This is why most production systems pair the LLM with a retrieval layer that pulls from a verified, up-to-date knowledge source before generating a response.

**Do I need to train my own LLM to use one in my business?**
No. Most businesses access LLMs through an API or a pre-built platform. The important configuration step is connecting the model to your own content - product catalogs, FAQs, policy documents - so it answers from your data rather than from general training knowledge.