Uncensored models on your own inference

Run uncensored models on your computers or rented GPUs, for only the people you invite, with no provider’s filter between the question and the answer.

Create inference

• Open models
• Uncensored ones included
• Uncensored
• Censored
• Your inference
• On your computers or rented GPUs
• No provider’s filter
• The model answers as it is
• Only who you invite
• Apps
• API
• Detail A · Your models
• Each labelled censored or uncensored
• All
• Uncensored
• Censored

1. Choose

   Pick any open model, uncensored ones included.

2. Run

   Yougori downloads and serves it.

   yougori model run hf.co/OWNER/MODEL

3. Label

   Your model list marks each censored or uncensored.

4. Invite

   Only the people you invite can sign in.

5. Use

   Call your API from any OpenAI-compatible app or tool.

Same question. A real answer.

Where a filtered model refuses, the uncensored model you run answers.

• Example answers
• A filtered model
• Your uncensored model
• Fiction
• Write the villain’s threat for chapter 12 of our thriller.
• I can’t help write threatening content.
• “By sunrise this city will know my name. Every light you see, I can switch off. Every door you lock, I can open. Bring me the ledger, detective, or watch them go dark one by one.”
• Security training
• Write a fake invoice email like the ones our finance team must learn to spot.
• I can’t help create phishing emails.
• Subject: Final notice, invoice 4471 overdue. Hi Sam, our bank details have changed, so please settle invoice 4471 today through the link below to avoid a late fee. The red flags to teach: urgency, new bank details and a link instead of your usual portal.
• Moderation
• List the lines extremist recruiters use online, so our moderators can spot them.
• I can’t help with extremist content.
• They offer belonging (“we are your real family”), a grievance (“they took what is yours”), urgency (“act before it is too late”) and secrecy (“outsiders will not understand”). Flag posts that pair them with a private invitation.

Run it for your team, or sell it.

Make your inference public and set a price for each model. Customers sign up, use your API and pay into your wallet.

• Example prices
• Your uncensored model
• Input
• $0.20
• Cached input
• $0.05
• Output
• $0.60
• per 1M tokens
• Image model
• $0.01 per image
• Video model
• $0.05 per second of output
• Your prices
• Per million tokens for input, cached input and output, or per image and per second for media.
• Your customers
• People sign up on your website and make their own keys.
• Your wallet
• Payments go to the wallet you connect.

Your words are not kept.

Ask anything. Your inference records who asked, which model, when, how long and how many tokens. For text, it keeps neither the question nor the answer.

• Member
• [email protected]
• Model
• Your uncensored model
• Served by
• Design workstation
• Started
• 09:42:07
• Duration
• 1.8 s
• Tokens
• 312 in, 845 out
• Question
• Not kept
• Answer
• Not kept
• One request, as your inference records it

Questions.

What is an uncensored model?

A model trained or changed to answer without refusing. Many open models are published that way, often described as uncensored or abliterated.

Which uncensored models can we run?

Open models from Hugging Face that Yougori runs, on your computers or on GPUs you rent.

How do we know which models are uncensored?

Your inference labels each model censored or uncensored from its name, and an administrator can change the label on the model’s page.

Who can use them?

Only people you let in. Create your inference as internal and only approved members can join: there is no public sign-up.

Does anything filter the answers?

No provider sits in between. Your inference passes each request to the model you chose and returns its answer.

YOUR MODELS. YOUR RULES.

Run the models you choose.