> For the complete documentation index, see [llms.txt](https://docs.roboflow.com/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.roboflow.com/universe/find-a-model-on-universe.md).

# Find a Model on Universe

Over 50,000 datasets on Universe also have trained models associated with the dataset. These models can be used in your computer vision project with [Workflows](/workflows/what-is-workflows.md).

To find a dataset on Universe, type in what you want to identify into the search bar. On the search page, check the "Has a Model" box:

<figure><img src="/files/wU3FbDzu8HyfBhqestIp" alt=""><figcaption></figcaption></figure>

This will refine your search to only datasets that contain trained models.

Click on a result to find out more about the project.

The dataset home page will show metrics. These metrics will depend on the model type. For object detection, for example, you will see mAP, precision, and recall; for classification you will see accuracy.

<figure><img src="/files/hFaTpOt933T4t5h3zi5L" alt=""><figcaption></figcaption></figure>

You can upload an image into the "Try This Model" section to try the most recent model version.

You can also click "Models" in the sidebar to see all trained models and choose one to try:

<figure><img src="/files/KYhFosCQIMvcZTm0qrkx" alt=""><figcaption></figcaption></figure>

When you open the Models page, you will see the results from an example image in the project test set plotted onto the image:

<figure><img src="/files/oBtqNqNBQPdfJd6EOGVA" alt=""><figcaption><p>The class "occupied forklift" is identified in the image.</p></figcaption></figure>

## Deploy a Universe Model

Each Universe model page has a "Deploy Model" button that opens a modal with several deployment options:

* Code Snippet: ready-to-use code in Python, JavaScript, or cURL for calling the model via the [Serverless API](/deploy/serverless-hosted-api-v2.md). Each snippet includes a "Copy prompt" button that copies a structured prompt you can paste into an AI coding agent.
* MCP: setup instructions for connecting the [Roboflow MCP Server](/agents/agents.md) to Claude Code, Cursor, or Codex. Once connected, your agent can run inference on Roboflow models directly.
* Custom Workflow: pre-built [Workflow](/workflows/what-is-workflows.md) templates for common tasks like detection, counting, or visualization.

You can also type a prompt directly in the Roboflow Agent input at the top of the modal to start a guided conversation about building with the model.

### Use a Universe Model in a Workflow

You can use Universe models in Workflows.

To use a model, open Roboflow Workflows, [create a Workflow](/workflows/create-a-workflow.md), then add a model block:

<figure><img src="/files/pF46X8zLs8fXhggGc99Q" alt=""><figcaption></figcaption></figure>

Click "Model" in the block editor to open the model selector:

<figure><img src="/files/vEe3NmuWH6wasadsq4rH" alt=""><figcaption></figcaption></figure>

Copy the model ID from the top of the Universe model page, then type or paste it into the model selector's search box. Model IDs have the form `workspace/project` or `workspace/project/version`.

Roboflow checks the ID against your workspace and Universe, then fills in the model's name, classes, task type, and sample images. Select the matching result to use it. If the ID does not resolve to a real model, a "not found" message shows and the model cannot be selected.

The same model selector is used in the deploy flow, so you can enter a custom model ID anywhere you pick a model.