# Baseline Model

> A very simple model you build first. Any real model has to beat its score.

**Field:** Artificial Intelligence  
**Category:** Evaluation & Benchmarks  
**Also known as:** Baseline, Naive Baseline

## Think of it like

It is like timing a walk before you try running. Then you know if running is faster.

## Definition

A baseline model is a very simple model you build first, so you have a score to beat. It might always guess the most common answer, or follow one plain rule. Any new model has to do better than it to be worth the work. A baseline also catches a trap. If 95% of emails are not spam, a model that says 'not spam' every time is 95% right and of no use.

## Example

Emma's baseline for house prices just guesses the average price. Her new model's guesses are off by 8%, the baseline's by 25%, so the new model earns its place.

## Why it matters

A score means little on its own. The baseline tells you whether it is good or just looks good.

## Common mistake

Skipping the baseline and calling a score of 90% good. On some tasks, a plain guess can reach 90% by itself.

## Three principles

- **Simple first.** One plain rule.
- **A score to beat.** For every new model.
- **Catches traps.** Easy tasks look hard.

## Related terms

- [Model Evaluation](https://learn.benibix.com/term/ai/model-evaluation)
- [Accuracy (Machine Learning)](https://learn.benibix.com/term/ai/accuracy)
- [Regression (Machine Learning)](https://learn.benibix.com/term/ai/regression)
- [Classification (Machine Learning)](https://learn.benibix.com/term/ai/classification)

![Two bars for error. The baseline that always guesses the average is off by 25%. The new model is off by 8%, well below it.](https://learn.benibix.com/diagrams/ev-baseline-model.svg)

---

Source: https://learn.benibix.com/term/ai/baseline-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: 2026-11-19
Licence: CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). Quote freely with attribution.
