---
title: "ARX for model providers"
description: "How ARX deploys your models: the interface path from a Hugging Face repository to a running model instance, the engines it supports, and the API your models answer on"
canonical_url: "https://developer.ardge.com/en/guide/model-provider/overview"
---
# ARX for model providers

> How ARX deploys your models: the interface path from a Hugging Face repository to a running model instance, the engines it supports, and the API your models answer on

You build and tune models. ARX is the platform that runs them on the customer's own hardware and puts them in front of the applications installed on the same system.

## What ARX handles

- **The deployment path** — AI Model Center pulls the model from your Hugging Face repository, downloads the weights, and starts the instance. You follow progress on the **Downloaded Model** and **Model Instance** pages
- **The inference engine** — vLLM, NVIDIA vLLM, Llama.cpp, and Faster Whisper are available; you choose the engine and its parameters while deploying
- **GPU capacity** — ARX estimates what the model needs and blocks a deployment that would not fit on the target card
- **The API surface** — a running model answers on the ARX model gateway in both OpenAI and Claude API format, so applications on the system call it without a vendor-specific client

## What you provide

- A model in **safetensors** or **GGUF** format, with the configuration and tokenizer files
- A Hugging Face repository to publish it from — private repositories work once an access token is saved in ARX
- The deployment settings you want users to start from

## Guides

<card-group>
<card icon="i-lucide-upload-cloud" title="Deploy your fine-tuned model on an ARX device" to="https://developer.ardge.com/en/guide/model-provider/deploy-fine-tuned-model">

Check your model against the formats and files ARX requires, upload it to a private Hugging Face repository, then download and deploy it from AI Model Center

</card>
</card-group>

<note>

**Calling a deployed model:** Once a model is running, applications reach it through the model gateway. See [Call on-premises models on ARX through the API](https://developer.ardge.com/en/guide/isv/api-model-calls) for the endpoint, key, and model ID.

</note>


## Sitemap

See the full [sitemap](https://developer.ardge.com/sitemap.md) for all pages.