Choose a model
Artemis Desktop connects to a locally running large language model (LLM) through an API endpoint. You are responsible for downloading and preparing your chosen model, installing a serving engine, and starting the model server. Artemis Desktop connects to the running model but does not install or manage it.
Setting up a model takes three steps:
- Choose a model and a serving engine (this page).
- Run the model server with OpenVINO or with llama.cpp.
- Connect the model to Artemis Desktop.
Recommended models
| Model | Notes |
|---|---|
| Qwen3 Coder 30B A3B Instruct | Recommended default. |
| Qwen3.6 35B A3B | Alternative recommended model. |
Choose a model based on your hardware and performance requirements.
Local models are limited by the memory and compute resources available on your machine. Larger or more capable models, and longer context lengths, need more memory.
Recommended serving engines
A serving engine loads the model and exposes it through an OpenAI-compatible API endpoint that Artemis Desktop can connect to.
Both recommended models can be served with either engine. Choose the engine that suits your hardware.
| Engine | When to use it |
|---|---|
| OpenVINO | Supported hardware configurations, particularly Intel AI hardware. See Run with OpenVINO. |
| llama.cpp | Serving the models as GGUF files on most other hardware. See Run with llama.cpp. |
Other OpenAI-compatible servers, such as Ollama, LM Studio, vLLM and Docker Model Runner, can also be added as a Self-Hosted source.