> For clean Markdown of any page, append .md to the page URL.
> For a complete documentation index, see https://docs.nvidia.com/sdgm/llms.txt.
> For AI client integration (Claude Code, Cursor, etc.), connect to the MCP server at https://docs.nvidia.com/sdgm/_mcp/server.

# Kumo Relational NIM

> Use Kumo Relational as an NVIDIA NIM inference service through the Kumo Relational Client.

[Kumo Relational](https://arxiv.org/abs/2604.12596) is a pretrained foundation
model for predictions across related tables. It uses examples supplied with a
request rather than requiring a separate trained model for each prediction task.

The **Kumo Relational NIM** hosts the model for inference. Use
[Kumo Relational Client](/client/overview) to submit prediction requests to its
endpoint and receive results in your Python application.

## What you can predict

Choose the prediction type that matches the output you need:

#### [Classification and Regression](/kumo-relational/user-guide/make-predictions/prediction-types)

Predict categories, Boolean outcomes, and numeric values.

#### [Forecasting](/kumo-relational/user-guide/make-predictions/forecasting)

Predict numeric values over a future time horizon.

#### [Link Prediction](/kumo-relational/user-guide/make-predictions/prediction-types)

Rank or recommend related entities.

## Use the NIM

* [Deploy, Install, and Connect](/kumo-relational/get-started/sdk-getting-started)
  covers endpoint setup, client installation, and authentication.
* [Quickstart](/kumo-relational/get-started/rfm) walks through a first prediction.
* [How Kumo Relational Works](/kumo-relational/about-kumo-relational/introduction)
  explains how related records and labeled examples become model input.