# Run local LLMs on the Ryzen AI NPU with Lemonade Server

How-to · Local AI · 2 min read. By [David Wilson](https://testbenchlab.com/authors/david-wilson/). Updated 2026-10-02.

Lemonade provides local model tools for several backends, including supported Ryzen AI routes. Selecting Lemonade does not automatically mean the NPU is active: the model and backend determine execution.

**Requirements:** compatible Windows hardware, Ryzen AI drivers and a model supported by the intended backend. This is a documentation-based procedure, not a hardware validation claim.

## What is Lemonade Server?

Lemonade organises model execution behind an application and service interface. A model using llama.cpp is different from one prepared for a Ryzen AI NPU backend, even when both appear in the same application.

## Install Lemonade Server

Download and run the current official Windows MSI installer. Update Ryzen AI drivers as directed by the compatible AMD guide. Open the application and check the detected hardware and backends. Use the current installer rather than an executable name from an older tutorial.

## Run a model on the NPU

Select a catalogue model explicitly supported by the detected Ryzen AI backend, download it and start a chat. Inspect backend status and logs. A response proves that a model ran, not that it used the NPU.

## NPU vs iGPU in Lemonade

Choose the route according to model support. The current command interface can be checked with:

```powershell
lemonade --help
lemonade status
lemonade backends
```

A hybrid path can use multiple processors and should be labelled accordingly. Speed comparisons need matching model conditions.

## Frequently asked questions

### Does Lemonade Server use the NPU?

It can with compatible hardware, drivers and a supported model/backend. Other models use GPU or CPU routes.

## Related reading

- [What is an NPU, and does your PC need one?](https://testbenchlab.com/guides/what-is-an-npu/)
- [NPU vs GPU: which runs AI better on a mini PC?](https://testbenchlab.com/guides/npu-vs-gpu/)
- [How to install Ollama on Windows 11 (Ryzen AI mini PC)](https://testbenchlab.com/guides/install-ollama-windows-11-ryzen-ai-mini-pc/)

## Sources

1. [Lemonade installation documentation](https://lemonade-server.ai/docs/guide/install/)
2. [AMD Lemonade getting-started guide](https://developer.amd.com/playbooks/lemonade-getting-started/)
3. [Lemonade current command interface](https://lemonade-server.ai/docs/guide/cli/)