> ## Documentation Index
> Fetch the complete documentation index at: https://runpod-b18f5ded-lg-modelrepotest.mintlify.site/llms.txt
> Use this file to discover all available pages before exploring further.

# Overview

> Store, version, and pre-cache your model files on Runpod infrastructure.

<Note>
  Model Repo is currently in beta and is available on Mac and Linux only. Windows support is coming soon.
</Note>

Model Repo is a private model storage service built into Runpod. Upload your model files once, and Runpod caches them directly on your Serverless worker hosts so they are ready before the worker starts. No external download is required at cold start time.

## Key benefits

* **Faster cold starts**: Models are pre-cached on worker hosts when available, so workers start faster without waiting for an external download.
* **No HuggingFace dependency**: Your models are stored in Runpod's infrastructure, so endpoints don't require an outbound download on every cold start.
* **Private storage**: Models are stored in your account and are not accessible to other users.
* **Version control**: Each upload is content-addressed, so you can pin an endpoint to an exact model version and roll back at any time.

## How it works

```mermaid theme={"theme":{"light":"github-light","dark":"github-dark"}}
flowchart LR
  A["Upload model\nrunpodctl model add"] --> B["Runpod hashes\n& stores files"]
  B --> C["Configure endpoint\n--model-reference URL"]
  C --> D["Worker host\npre-caches files"]
  D --> E["Worker starts\nfiles ready on disk"]
```

1. Upload your model files using `runpodctl model add`.
2. Runpod computes a content hash and stores the files in private, secure storage.
3. Reference the model using the URL `https://local/{user-id}/{model-name}:{hash}` when configuring your endpoint.
4. When a worker starts, Runpod pre-caches the model files on the host before the container boots.
5. Your handler reads the model from a local path inside the container.

## What you can upload

Model Repo accepts any file format — PyTorch checkpoints, GGUF, safetensors, ONNX, or any format your worker needs. There is no type checking.

* **Max file size:** 5TB per file
* **Retention:** Files are retained indefinitely
* **Total storage limit:** Still being finalized

## Model Repo vs network volumes

**Use Model Repo when:**

* You have fixed, versioned model weights and want faster cold starts without external downloads
* You need version control over exactly which checkpoint is deployed
* Your endpoint spans multiple datacenters — Model Repo works across all DCs, while a network volume is tied to one

**Use a network volume when:**

* Workers need shared read/write access during a run (e.g., checkpoints, fine-tuning outputs)
* Your files change frequently and all workers need to see updates immediately
* You need many concurrent workers accessing the same files simultaneously

## Your data

Your models are stored in private, secure storage and are isolated to your Runpod account — no other users can access or list your models. Runpod does not access, analyze, or use your model files for any purpose.

For encryption details, certifications, and compliance information, see [Security](/serverless/storage/modelrepo/security).

## Getting started

See [Model Repo testing](/serverless/storage/modelrepo/testing) for a step-by-step guide to uploading a model and deploying it to a Serverless endpoint.