Home / request a model

Give your model an API.

Tell Umbra what your public Hugging Face model does and who needs it. We will review it manually, then can surface it to our Apple Silicon provider network.

Public model weights only. Internal or private API use is fine, but the model weights must already be public on Hugging Face. Do not submit Hugging Face keys, private weights, prompts, user data, proprietary training data, or other secrets.

Identify the model

Public Hugging Face repositoryUse the repository root. A public GGUF build is the fastest path to provider review.

Your name

Your relationship to the model

State what it is for

Primary use

Internal or private workloadsA company, team, or personal workflow will call the API.Public testing and discoveryYou want more people to try the model cheaply.Product or application backendThe model will power a feature, agent, or customer-facing app.Research or evaluationYou need repeatable remote access for experiments.Another useExplain it in the summary below.

Who should be able to call it?

What does the model do, and why does it need an API?

Describe the hosting gap

Current situation

Expected demand (optional)

Give us a way to follow up

The model weights are public on Hugging Face. I understand Umbra does not accept private weights or buyer uploads.Umbra may contact me about manual review, clarification, provider interest, and whether the model becomes available. See the Privacy Policy.

What happens after you submit

Manual review

We verify the public repository, intended use, license posture, model format, and whether the architecture fits Umbra.

Provider signal

An approved request can be promoted in the provider workbench with your use case and expected demand, but without your contact details.

Normal onboarding

An interested provider opts in, pulls the public model with their own Hugging Face key, and passes the same verification and attestation path as every other model.

Before you request

Can I request a private or gated Hugging Face model?

No. Umbra only hosts public model weights. The API usage can be for an internal team, but the weights themselves must be public on Hugging Face.

Does an approved request force providers to host the model?

No. Approval makes the request visible as an operator-reviewed demand signal. Each provider still chooses the models they host, and normal compatibility, licensing, integrity, and onboarding checks still apply.

Does submitting guarantee that the model will be listed?

No. Umbra reviews each request manually. Unsupported architectures, unclear licenses, non-public weights, and models that do not fit the network may be declined or held for more information.

Give your model an API.

Request received.

What happens after you submit

Before you request