Answers, not lectures
Abliterated models have their built-in refusals removed. You ask, they answer.
PRIVATE BETA · NEW UNCENSORED MODELS, COVERED DAILY
Intelligence, freed.
Uncensored, abliterated AI.
Pick a model that actually answers. We run it for you on a private cloud GPU, or set it up on your own machine, and you use it from the apps you already have. No giant downloads. No servers to babysit.
Open source · Private access · Real people, not a ticket queue
LATEST UNCENSORED RELEASES
The newest models that actually answer.
Every week brings new uncensored releases. We read the sources, check the numbers and write it up in plain language. Like one? We get it running for you.
4.1B/1.7B · HYBRID SWA · 1M CTX · APACHE-2.0 · DAY-TEN UNCENSORSsoyaakinohara/
apetersson/
0xSojalSec/
soyaakinohara/Spark-X2.5-4B-Heretic
Day ten for Spark X2.5: the uncensor wave on the 1M-context 4B
Read the review → 285B-A13B · VISION MOE · FIRST MULTIMODAL V4 · 11-HOUR ABLITERATIONapetersson/DeepSeek-V4-Flash-Vision-Exp-Abliterated
Eleven hours from DeepSeek drop to uncensor.
Read the review → 30B-A3B · TRANSLATION MOE · HERETIC ARA0xSojalSec/Tencent-Hy-30B-A3B-uncensored-heretic
Why would a translation model refuse? Tencent's Hy-MT2, decensored
Read the review →HOW IT WORKS
Three steps. Nothing to install.
Say hi on Signal
Tell us which model you want and what you'd like to do with it.
We set it up
On a private cloud GPU we rent for you, or on your own hardware. You get the price before anything starts.
Use it in your apps
Works with the chat and coding tools you already use. Keep it running, or switch it off any time.
WHY ABLITERATED.CLOUD
Your model. Your GPU. Your rules.
Private by design
Your model runs on a GPU that is yours alone, reached over a private connection. No shared chatbot.
Free, open code
Everything we build is MIT-licensed. Self-host it yourself whenever you like. No lock-in.
A real person helps
Setup, questions, switching models: message a human on Signal, not a ticket queue.
WHAT IT COSTS
No subscription. No surprise bills.
Our code is free. You pay for the GPU hours you actually use, plus the setup help you want, agreed up front.
| What you use | What it costs |
|---|---|
| One hour, while it runs | about $0.63 |
| A full day, non-stop | $15.20 |
| About two hours a day, for a month | $60.00 |
| Switched off, model kept ready | $24.00 a month |
You only pay while it runs. A switched-off machine keeps paying for storage until you delete it. Setup help is priced separately, before we start. Taxes and other services are extra.
Model weights are free to download but keep their own licenses. Check them before commercial use.
PLAIN ANSWERS
Questions people ask.
What does abliterated mean?
An abliterated model had its built-in refusals removed, so it answers instead of lecturing. It can still make mistakes, and no model says yes to everything.
Is it free?
Our code is free and open source. GPU time and setup help cost money, agreed up front. There is no subscription.
Do I need my own hardware?
No. We rent a private cloud GPU for you. If you already own a strong GPU, we can set it up there instead.
Which apps work with it?
Any app that talks to OpenAI-style models: chat apps, coding assistants, agent tools and LLM routers. We help you connect them.
Can I get an API key right now?
Not yet. Access is private and by invitation. Message us on Signal and we set you up personally.
PRIVATE BETA
Try the model you want. Know exactly what is running.
Tell us the model, the app and what you'd like to try. We reply on Signal.