> 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.

# How are Kumo foundation models different from an LLM or text-to-SQL?

> Distinguish prediction from language generation and database retrieval

Kumo foundation models predict outcomes from labeled examples and structured data without task-specific training. Kumo Tabular works with a single table; Kumo Relational also uses related tables.

An LLM generates or interprets language. Text-to-SQL translates a request into SQL that retrieves or aggregates existing records when executed.

| Question                                               | Typical approach                              |
| ------------------------------------------------------ | --------------------------------------------- |
| How many orders did this customer place last month?    | SQL, written directly or generated from text. |
| Summarize this customer's support conversations.       | An LLM.                                       |
| Will this customer place an order in the next 30 days? | A predictive model using historical data.     |

An application can combine an LLM to interpret requests, SQL to retrieve data, and a Kumo model to make predictions.

See [Choose a Model](/about-nvidia-kumo/overview#choose-a-model) for the available foundation models and [Kumo Relational](/kumo-relational/about-kumo-relational/overview) for its prediction workflow.