Providers
Uncensored models
Access to models with reduced refusals, what is moderated, and how suspensions and appeals work.
このページはまだ翻訳されていません。英語版を表示しています。
Uncensored models answer a wider range of questions than safety-trained models. They are served on Experiential Cloud, available to account holders outside a few sanctioned countries, and subject to content moderation on a narrow set of abuses. Your data controls, including zero data retention, apply as for any model.
How to recognize one
Each uncensored model's page opens with a content moderation notice, and the listing marks it. A request to any other model never falls back onto an uncensored one.
| Example slugs | |
|---|---|
deepseek-v4.1-flash-uncensored | |
glm-5.3-uncensored | |
glm-5.3-flash-uncensored |
{
"id": "glm-5.3-uncensored",
"object": "model",
"uncensored": true
}Access
| Use | Requirement |
|---|---|
| Text | An account outside the restricted countries; no KYC |
| Image input | KYC review |
| Early restoration of suspended access | KYC review |
| Zero data retention | Pro, standard provider rules, no separate approval |
Not available from Iran, North Korea, Cuba, Syria, Russia or Belarus, or to organizations registered there: 451 model_unavailable_in_region. For KYC, email founders@experientiallabs.ai from your work address with your organization name and ID, your use case and proof of business. Every request is reviewed by hand.
{
"model": "glm-5.3-uncensored",
"provider": { "zdr": true },
"messages": [{ "role": "user", "content": "..." }]
}Or send the X-Gateway-ZDR: true header. See Data controls.
What is moderated
| Category | Covers |
|---|---|
| Child sexual exploitation and abuse | Any sexual content involving minors, grooming, sharing such material |
| Mass-casualty violence and weapons | Attack help; chemical, biological, radiological, nuclear or mass-casualty explosive weapons |
| Fraud and extortion | Scams, theft by deception, identity theft, laundering, counterfeiting, blackmail |
Moderation runs on our own classifiers without storing prompts or responses. A flag is an automated assessment and can be wrong; a response is not a promise that a request complied.
| Not moderated | |
|---|---|
| Sensitive subjects in general | Allowed |
| Topics a safety-trained model would refuse | Allowed, outside the three categories |
Suspensions and appeals
Repeated flags suspend your organization's uncensored access; every other model keeps working. Each new suspension lasts longer, and only new requests after one ends can trigger the next.
| Suspension | Length |
|---|---|
| 1st | 1 hour |
| 2nd | 24 hours |
| 3rd | 7 days |
| 4th | 30 days |
| 5th | Permanent |
Appeal any suspension, permanent included, in Settings > Organization, or by email to founders@experientiallabs.ai with your organization ID and request IDs. Resellers can scope suspensions to one end customer: see Resellers.
{
"error": {
"type": "permission_error",
"code": "uncensored_access_suspended",
"message": "..."
}
}