Request Access to a Privileged Model

Privileged models need an approved access request before your team can call them. Request access from the dashboard, track the request status, and accept updated terms.

A privileged model has reduced or removed safety behavior, or is otherwise high-risk. Your team must be approved before it can call one. For the list of privileged models and the policy summary, see Privileged Access.

Approval is per team and per model. An approval for one model does not give access to another.

Find a privileged model

A privileged model shows a lock badge (Requires approval) in the model catalog, on its model page, and in the playground model picker. Its model page shows a banner, This model requires approved access, with your team's status and a Request access button.

In the API, GET /v1/models marks the model with "requires_approved_access": true and adds your team's access_status. See Check access in GET /v1/models.

Request access

Open the model page

Sign in, select the team that will call the model, and open the model page. The 403 error from the API also links to the request page, https://inference.net/model-access/<model-id>.

Click Request access

Fill in the form:

Field	Required
Legal name (the person or organization responsible for the team)	Yes
Country	Yes
Intended use	No

Check the box to confirm that you are 18 or older, that you are authorized to act for the team, and that you accept the Privileged Model Access Terms.

Wait for the decision

We review every request. You get an email when your request is approved or rejected. Approved access reaches the API within 45 seconds.

The Privileged Model Access Terms are on each model's request page, https://inference.net/model-access/<model-id>. The current version is 2026-10-23.

Access status

`access_status`	Dashboard badge	Meaning	What you can do
`none`	Requires approval	Your team has not requested access.	Request access.
`pending`	Access pending	The request is under review.	Wait for the email.
`approved`	Access approved	Your team can call the model.	Call the model.
`rejected`	Access rejected	The request was rejected.	Request access again.
`revoked`	Access revoked	An earlier approval was revoked.	Request access again.
`terms_required`	Access paused	Your team was approved, but on terms that no longer grant access. Calls fail until you accept the current terms.	Accept updated terms.

Only approved lets your team call the model.

Accept updated terms

When the terms change, the model page shows Accept updated terms. Click it, check the box, and click Accept terms. The form keeps your earlier answers, and you cannot change them.

For a minor change, your access continues while you have not accepted. The status stays approved.
For a major change, access stops until you accept. The status is terms_required.

Access comes back within 45 seconds of acceptance.

Propagation

Change	Time to reach the API
Approve, revoke, or terms acceptance	Up to 45 seconds
`access_status` in `GET /v1/models`	Up to 6 minutes

Check access in `GET /v1/models`

Send your API key to get your team's status. Without a key, access_status is always none.

curl -s https://api.inference.net/v1/models \
  -H "Authorization: Bearer $INFERENCE_API_KEY" \
  | jq '.data[] | select(.requires_approved_access) | {id, access_status}'

{
  "id": "deepseek-v4.1-flash-1017-abliterated",
  "access_status": "pending"
}

The `403 model_access_required` error

A call to a privileged model without approval returns HTTP 403:

{
  "error": {
    "message": "Model 'deepseek-v4.1-flash-1017-abliterated' requires approved access. Request access at https://inference.net/model-access/deepseek-v4.1-flash-1017-abliterated",
    "type": "model_access_required",
    "code": "model_access_required"
  }
}

Match on error.type. Some responses do not include code.