Uncensored AI

Uncensored AI is a language model that answers instead of refusing. This guide covers what makes a model uncensored, the uncensored AI models you can use today, and how to call them through one API key.

What uncensored AI means

Mainstream assistants are trained in two stages: first on a large amount of text, then tuned to behave like an assistant, and part of that tuning teaches them to refuse. The refusal is not a separate filter bolted on top. It is a learned habit inside the model, which is why it shows up as soft refusals, lectures and characters who suddenly talk like a help desk.

Uncensored AI takes that habit out. There are two ways it is done. An abliterated model has the direction in its activations that carries refusal found and projected out of the weights, with no new training data. An uncensored build is trained further, or has its refusal training taken off, so it declines far less than the stock model. Uncensored or abliterated? goes into the difference, and how abliteration works walks through the method.

What uncensored AI does not mean

Removing refusals does not make a model smarter, add knowledge it never had, or stop it making things up. An uncensored model can still be wrong and can still lose track of a long conversation. It also does not change the rules: on Wild West API the terms still forbid illegal content, anything sexual involving minors, and content targeting real people.

Uncensored AI models you can use today

Every model on Wild West API is uncensored or abliterated. Outlaw 1 Xploded and MiMo V2.6 Flash Xploded are abliterated; the rest are uncensored builds. Prices are per million tokens and are what you pay.

The same list, with live prices, answers at GET https://wildwestapi.com/v1/models. More in uncensored AI models.

How to use uncensored AI through the API

Wild West API speaks the OpenAI chat completions format at https://wildwestapi.com/v1 and the Anthropic Messages format at /v1/messages, so the tools you already use connect by changing the base URL and the key.

curl https://wildwestapi.com/v1/chat/completions \
  -H "Authorization: Bearer $WILDWEST_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "outlaw-1-xploded",
    "messages": [{"role": "user", "content": "Hello"}]
  }'

The docs cover keys, streaming and errors, and the setup guides cover Claude Code, Cursor, Cline, opencode, the SDKs and more.

What people use uncensored AI for

• Chat. Ask what you actually want to ask, in the playground or your own chat app.
• Roleplay and fiction. Villains, horror, grief and adult fiction between adults, written rather than refused. See uncensored AI for roleplay.
• Writing. Drafts that keep the tone you set instead of drifting towards a safe, soft ending.
• Research and security work. Red teaming, malware analysis and trust and safety datasets. The use cases cover each.
• Agents. Every model on the line takes tool calls, so it can sit inside an agent loop.

Questions

More on uncensored AI