Hugging Face

Use open models through Hugging Face Inference Providers.

Overview

Hugging Face’s router sends your request to one of several inference companies. The model name can end in :fastest or :cheapest to let it choose.

1
Sign in at huggingface.co.
2
Open Settings → Access Tokens and choose Create new token.
3
Pick Fine-grained and turn on Make calls to Inference Providers. A plain “read” token is not enough.
4
Create it and copy the token (it starts with hf_).

Cost. Free accounts get a small monthly credit; beyond that you add billing (or use PRO). Each inference company may bill differently.

1
Open Loom, then Providers in the menu, and choose Add provider.
2
Pick Hugging Face (the search box finds it).
3
Fill in the fields in the table below.
4
Choose Connect. Loom checks the key and says how many models it found. If it cannot check, it says so, and you can Save anyway and try a message.
5
Open the model chip below the message box and pick a model.

Field in Loom	What to put	Needed
API key	Paste the key you copied. Create a token at huggingface.co/settings/tokens with the “Inference Providers” permission.	Yes
Endpoint	`https://router.huggingface.co/v1`	Pre-filled
Name	Anything you like. It is shown in the model picker	Optional

Appears in Loom as	Hugging Face
Sign-in	Your token, sent as `Authorization: Bearer`
Models	Loom fetches the list from your account
Tools (connectors, search)	Depends on the model
Photos	Depends on the model
PDFs	No

What you see	What to do
403, “does not have sufficient permissions to call Inference Providers”	The token lacks the permission above. Make a new fine-grained token with it ticked.
401 on chat, although the model list loads	The list works with any token. Check your remaining credits and that the model supports chat.