# Hardware

Tell the Hub which compute hardware you own (GPUs, CPUs, or Apple Silicon) and it will help you find models that run on your setup. It also appears on your public profile, so you can compare your setup with the community, and you can make it private any time.

Manage it any time from your [Hardware settings](https://huggingface.co/settings/hardware).

## Add your hardware

On your [Hardware settings](https://huggingface.co/settings/hardware) page, add each piece of hardware you own:

1. Pick a **type**: GPU, CPU, or Apple Silicon.
2. Choose the **provider** and **model** (for example, NVIDIA RTX 4090).
3. Set the **memory** (VRAM, RAM, or unified memory) and the **number of units** you have.

Add as many items as you like. If you own several, mark one as your **primary** hardware so it appears first on model pages.

> [!TIP]
> Your hardware is public by default. Turn off the **Publicly Visible** toggle to keep it private.

## See which models fit your hardware

On model pages that offer [GGUF](./gguf) or MLX files, a **Hardware compatibility** panel estimates whether each quantization will run on your saved hardware, so you can choose a size that fits before downloading. Pair it with [Local Apps](./local-apps) to get up and running in a couple of clicks.

## Share and compare

While your hardware is public:

- A **TFLOPS** badge appears on your profile, summarizing your estimated total compute power.
- You can browse [what the community is running](https://huggingface.co/hardware) and see how your setup compares across GPUs, CPUs, and Apple Silicon.

## Read hardware from the API

If your hardware is public, it is also [available programmatically](https://huggingface-openapi.hf.space/#tag/users/GET/api/users/{username}/overview):

- `GET https://huggingface.co/oauth/userinfo` includes a `hardwareItems` claim when your app was granted the `profile` scope, so a [Sign in with Hugging Face](./oauth) app can read the signed-in user's setup without an extra call.
- `GET https://huggingface.co/api/users/{username}/overview` returns a `hardwareItems` array for any user whose hardware is public.

```bash
curl https://huggingface.co/api/users/{username}/overview
```

```json
{
  "hardwareItems": [
    { "sku": ["GPU", "NVIDIA", "RTX 4090"], "mem": 24, "num": 2, "isPrimary": true }
  ]
}
```

Each item contains:

- `sku`: a `[type, provider, model]` triplet.
- `mem`: the memory in GB (VRAM, RAM, or unified memory).
- `num`: the number of units.
- `isPrimary`: only present on the item marked as primary.

If the user turned off the **Publicly Visible** toggle, `hardwareItems` is omitted from both responses. See the [Hub API documentation](./api) for the full response schemas.

## Next steps

- [Use AI Models Locally](./local-apps) — run models with your favorite local app.
- [Local Agents with llama.cpp](./agents-local) — build a coding agent on your own hardware.