> ## Documentation Index
> Fetch the complete documentation index at: https://docs.labelbox.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Compare model runs

> Visualize predictions and compare metrics between model experiments.

As you iterate on your model, you'll want to compare the performance of different versions to see if your changes are having a positive impact. Labelbox makes it easy to compare two model runs side-by-side.

## Before you start

You will need a model with two or more model runs to use this feature. Follow the steps in this guide to learn how to create a model run.

<Card title="Create a model run" icon="plus" horizontal href="/horizon/guides/create-a-model-run" />

## How to compare two model runs

1. Go to the **Model** tab and select an Experiment.
2. Open one of the model runs you want to compare.
3. Click the **Compare against** dropdown and select the second model run.

After selecting two model runs to compare, you will be able to see the compared results overlayed on the thumbnails of the data rows.

### Compare visually

Once you have selected two model runs to compare, you can visually inspect the differences in their predictions. The predictions from each model run will be displayed in a different color, making it easy to see where they agree and disagree.

From the gallery, you can click on an individual data row to expand it. Using the right sidebar, you can toggle which ground truth annotations and predictions to view, along with which model runs to display in general. For even more visualization options, click **Display**.

### Compare performance metrics

You can also compare the performance metrics of two model runs. The confusion matrix, precision-recall curve, and scalar metrics will all be updated to show a side-by-side comparison of the two models.

You can compare both scalar metrics and confusion [metrics](/horizon/guides/model-metrics) between two model runs.

### Compare model run configs

To understand *why* your models are performing differently, you can also compare their configuration files. This will show you the differences in the hyperparameters used for each model run, which can often explain the differences in performance.