> For the complete documentation index, see [llms.txt](https://docs.layerlens.ai/llms.txt). Markdown versions of documentation pages are available by appending `.md` to page URLs; this page is available as [Markdown](https://docs.layerlens.ai/5.-select-pick-the-model/byok-custom-models.md).

# BYOK custom models

Register an OpenAI-compatible endpoint as a first-class custom model in Stratix Premium.

{% hint style="info" %}
**Available in Stratix Premium.** This surface is part of the logged-in workspace at [stratix.layerlens.ai](https://stratix.layerlens.ai). Stratix Public users can browse the catalog but cannot use this feature.
{% endhint %}

Stratix supports registering **OpenAI-compatible endpoints** as custom models. The custom model is treated as first-class: you can use it in evaluations, compare it against public models, score traces from it, and apply judges and scorers — exactly like a public model.

## When to use BYOK

* You're running a self-hosted model
* You're using a private inference endpoint
* You're using a partner cloud (Fireworks, Nebius, SambaNova, MLflow, etc.) with an OpenAI-compatible API
* You want to keep model API tokens billing under your own provider account, not as ECU

## Registering a custom model

1. **Premium → Models → Register custom model**
2. Provide:

* **Name** — display name for your team
* **Endpoint URL** — `https://api.example.com/v1` (an OpenAI-compatible base URL)
* **API key** — secret stored encrypted in your org's BYOK vault
* **Model identifier** — what to send as `model:` in the request
* **Capability tags** — what the model is good for

3. **Test connection** — Stratix sends a probe request and confirms the endpoint responds
4. **Save**

## What happens at evaluation time

When you select a BYOK custom model in an evaluation:

* Stratix sends inference requests to your endpoint
* Token usage bills against your provider, not against ECU
* Judges, scorers, and trace evaluations work exactly as for public models
* Latency, error rates, and cost-per-call are tracked in the evaluation results

## Compatible providers

Any OpenAI-compatible API works. Tested with:

* vLLM (self-hosted)
* Together AI
* Fireworks AI
* Nebius
* SambaNova Cloud
* MLflow gateways
* Custom endpoints behind a reverse proxy

If your provider isn't OpenAI-compatible, see if the SDK exposes a compatibility layer.

## Security

* API keys are encrypted at rest in the org's BYOK vault
* Keys are never returned via the API; only the masked prefix is shown
* Org-scoped: only members with the `models:write` permission can register or rotate keys

## Where to next

* [Concept: BYOK models](/5.-select-pick-the-model/byok-models.md)
* [Models](/5.-select-pick-the-model/models.md)
* [Settings](/11.-admin/settings.md)