API reference · Schema
ModelReference
objectstringrequiredAlways `model_reference`: the type of this object
One of
model_referenceidstringrequiredUnique ID (`mrf_…`)
ID with prefix mrf_
repositorystringrequiredThe repository on the model hub, as <owner>/<name>
requested_revisionstringrequiredThe branch, tag or commit the reference named when it was accepted
commitstringrequiredThe 40-character commit the revision resolved to
includearray of stringrequiredThe file patterns it was limited to (empty: every file)
visibilitystringrequiredprivate: fetched with token_secret_id and cached on hosts for this project only
One of
publicprivatetoken_secret_idstring | nullrequiredID of the token secret
file_countintegerrequiredHow many files it holds
total_bytesintegerrequiredThe size of all its files together, in bytes
Every file, with the size and digest each is verified against on the host
ShowHide 4 fields of
filespathstringrequiredPath of the file inside the repository
sizeintegerrequiredSize of the file in bytes
sha256string | nullrequiredLarge files: SHA-256 of the content
git_blob_sha1string | nullrequiredSmall files: the version-control object ID
mount_pathstringrequiredWhere it appears in a container: the cache layout the model hub client libraries read, so they find it with networking off
snapshot_pathstringrequiredThe directory holding its files inside a container
created_atstring (date-time)requiredWhen it was created (ISO 8601, UTC)
{
"object": "model_reference",
"id": "mrf_01jc5zq4m8k3v6x9t2w7r0b1e4",
"repository": "hf-internal-testing/tiny-random-gpt2",
"requested_revision": "string",
"commit": "string",
"include": [
"string"
],
"visibility": "public",
"token_secret_id": "string",
"file_count": 1,
"total_bytes": 1,
"files": [
{
"path": "string",
"size": 1,
"sha256": "string",
"git_blob_sha1": "string"
}
],
"mount_path": "/model-cache/models--hf-internal-testing--tiny-random-gpt2",
"snapshot_path": "string",
"created_at": "2026-01-01T00:00:00.000Z"
}Used by
- List the project’s model references
GET /v1/model-references - Accept a model reference (checked against the model hub, pinned to a commit)
POST /v1/model-references - Get a model reference and the files it resolved to
GET /v1/model-references/{reference_id}