# Base Model

> A model fresh from its first training. It carries text on but does not follow orders.

**Field:** Artificial Intelligence  
**Category:** LLMs  
**Also known as:** Pretrained Model, Raw Model

## Think of it like

Think of it like a parrot that has heard a lot. It goes on with what it hears, but it does not follow orders.

## Definition

A base model is a language model straight after its first big round of training. It has not yet been taught to follow orders. It is very good at carrying on text in the same style. Ask it a question and it may write more questions, as if it were filling in a list. Teams use base models as the starting point for chat models, and for research.

## Example

Jack types 'Write a poem about rain' into a base model. It replies with more lines like 'Write a poem about snow', because that is how such lists usually go on.

## Why it matters

Every chat assistant starts as a base model. That is why it needs more training to be of help.

## Common mistake

Testing a base model like a chat assistant and calling it broken. It was never trained to answer.

## Three principles

- **First round only.** Trained to guess the next token.
- **Carries text on.** In the same style.
- **A starting point.** For instruct and chat models.

## Related terms

- [Instruct Model](https://learn.benibix.com/term/ai/instruct-model)
- [Foundation Model](https://learn.benibix.com/term/ai/foundation-model)
- [Next-Token Prediction](https://learn.benibix.com/term/ai/next-token-prediction)
- [Pretraining](https://learn.benibix.com/term/ai/pretraining)

## Told apart from

- [Base Model vs Instruct Model](https://learn.benibix.com/compare/base-model-vs-instruct-model): People often think every language model is a chat assistant. A base model is only trained to carry text on. So a question may get more questions back. An instruct model is that same base model with more training to follow requests. The facts mostly come from the base. The manners come from the extra step.

## Taught in

- [Post-Training and RLHF](https://learn.benibix.com/learn/ai-llms/post-training-and-rlhf): LLMs, 8/15

![The same question goes to two models. The base model carries on with more questions in the same style. The instruct model answers the question.](https://learn.benibix.com/diagrams/llm-base-vs-instruct.svg)

---

Source: https://learn.benibix.com/term/ai/base-model
Author: Behnia Zarei (a developer in Vienna, Austria, who builds and teaches data, AI and business intelligence at BENIBIX)
Publisher: BENIBIX Learn (https://learn.benibix.com/)
Last updated: 2026-11-18
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
