# Deployment options

Bria's API runs in Bria's cloud by default, and that is the right choice for most teams. You can also run Bria models on your own hardware — self-serve with a per-model license, or through an Enterprise engagement for your own cloud account or data center.

| Option | Where inference runs | Who operates it | How to get it |
|  --- | --- | --- | --- |
| Bria Cloud | Bria's infrastructure (this API) | Bria | Just [call the API](/getting-started/quickstart) |
| Self-hosted license | Your own GPUs | You, with Bria's packages | [Buy a per-model license](/self-hosted/overview) — self-serve, by card |
| Enterprise & on-premises | Your cloud account or data center, including air-gapped | You, with Bria's packages and support | [Contact sales](https://bria.ai/contact-us) |


## Self-hosted licenses

Buy a license for a single model and run it on your own GPUs. It is self-serve (by card, no sales call), the model installs with [Bria Studio](/self-hosted/quickstart) or a manual install, and the weights come from Bria directly — no Hugging Face account. Usage shows in the Bria Console.

See [Self-hosted models](/self-hosted/overview) for the catalog, setup and pricing. Self-serve licenses are available to companies with revenue under $10M; above that, talk to [sales](https://bria.ai/contact-us).

## Enterprise and on-premises

An Enterprise engagement covers what a self-serve license does not:

- **On-premises and air-gapped** deployments in your own data center, and private networking inside your cloud account.
- **Full-precision and proprietary** model versions, with **automatic upgrades** to the newest models.
- An **SLA**, uncapped indemnity, and email plus priority support.
- **Fine-tuning and custom models** built by Bria, and the right to **resell or redistribute** outputs.
- A broader catalog than the self-serve models, with hardware guidance for your workloads.


Enterprise
On-premises and Enterprise deployments are scoped with Bria. [Contact us](https://bria.ai/contact-us) to plan one.

### How an Enterprise deployment works

1. **Provisioning.** Bria grants your organization access to a private package registry and, where needed, model weights, tied to your enterprise credentials — nothing else needs to be shared.
2. **Install and run.** Install the pipeline packages in your environment (GPU instances in your cloud account, or on-premises hardware) and run them from your own services or notebooks. Inputs and outputs never leave your environment.
3. **Licensing and updates.** Pipelines are licensed to your organization and use your Bria credentials to fetch updates.


## Attribution

Whenever Bria models run outside Bria's cloud — self-hosted license or Enterprise deployment — the [Attribution Agent](/enterprise/attribution) runs alongside them. It computes irreversible embeddings of generated outputs and sends only those vectors to Bria, which keeps every output attributed and covered by Bria's licensing terms without exposing your data.

## Model weights and Hugging Face

Licensed self-hosting fetches weights from Bria directly, so you never need a Hugging Face account. Bria's weights are also published on [Hugging Face](https://huggingface.co/briaai) (FIBO, FIBO Edit, RMBG 2.0 and more); they are source-available for non-commercial use, and commercial use requires a Bria license — a self-hosted license or an Enterprise agreement.

## Mixed deployments

You can combine options: for example, run a generation model on-premises for privacy-critical work while using Bria Cloud for other editing endpoints, or generate structured prompts with the hosted [`structured_prompt/generate`](/image-generation/generation/structured-prompt-generate) endpoint and render locally. Because [VGL](/vgl) structured prompts are portable across the Fibo family, assets and prompts move between environments without translation.

## Next steps

- [Self-hosted models](/self-hosted/overview) — buy and run a single model yourself
- [Attribution Agent and Service](/enterprise/attribution)
- [Safety and transparency](/safety)
- [Contact us](https://bria.ai/contact-us) to plan an Enterprise deployment