Skip to main content
LM Studio lets you download and run GGUF models through a local server with an OpenAI-compatible REST API. Like Ollama, it runs entirely on your machine — no data is sent externally.
LM Studio is free to use for local inference. Your Blueprint code and source files never leave your machine.

Setup

1

Download LM Studio

Go to lmstudio.ai and download the installer for your platform.
2

Load a model

Open LM Studio, search for a model (e.g., qwen3-32b), and download it. Once downloaded, load it into memory using the model selector.
3

Start the local server

In LM Studio, navigate to the Developer tab (or Local Server tab depending on your version) and click Start Server. The server listens on http://localhost:1234 by default.
4

Configure in Unreal Engine

Open Edit → Project Settings → Plugins → Node to Code → LLM Services → LM Studio and set:
  • Server Endpoint to your LM Studio server URL (default: http://localhost:1234)
  • Model Name to the model identifier shown in LM Studio (default: qwen3-32b)
5

Select LM Studio as your provider

Under Node to Code | LLM Provider, set Provider to LM Studio.

Configuration

All LM Studio settings are under Node to Code | LLM Services | LM Studio in Project Settings.
SettingDefaultDescription
Server Endpointhttp://localhost:1234The base URL of your LM Studio local server.
Model Nameqwen3-32bThe model identifier to request. This must match the model loaded in LM Studio.
Prepended Model Command(empty)Text prepended to the start of every user message.

Prepended model command

Some models support special commands that control their behavior. You can enter these in the Prepended Model Command field and the plugin will insert the text at the beginning of each user message automatically. A common example is /no_think, which disables extended thinking on reasoning models that support it:
/no_think
This can reduce latency significantly when using a reasoning model for straightforward translation tasks.
Check the documentation for the specific model you’re using to see what commands it supports. Not all models respond to /no_think or similar directives.

Build docs developers (and LLMs) love