Jailbroken AI
API access to the newest jailbroken frontier models, abliterated so they answer instead of refusing. One OpenAI and Anthropic compatible key, and you pay only for what you use.
Base URL https://wildwestapi.com/v1
Model said no? Get your money back. Email support@wildwestapi.com with the request and we'll refund what it cost.
New jailbroken models
New open frontier models with the refusal behaviour taken out, for reasoning, coding, research and agent work. We add them as they release.
Compatible endpoints
OpenAI and Anthropic formats, so the tools you already use connect by changing only the base URL and the key.
Advanced workloads
Red team evaluations, security research, cyber defence, coding agents and long roleplay, without a content filter in the way.
Red team with a model that does not refuse.
Mainstream APIs will not write the attack prompts, adversarial test cases or harmful samples a red team needs. An abliterated attacker model will, on a key with a hard spend cap.
Red team use cases
Penetration testing, jailbreak and prompt injection testing, malware analysis, detection engineering and more.
See the use casesRed team tools
garak, PyRIT, promptfoo and more, set up with Wild West API as attacker, judge or target.
Set up a toolCybersecurity playbooks
Phishing triage, Sigma and YARA rules, code review and incident timelines with a security-tuned model.
Open the playbooksUse it from your stack
LangChain, LlamaIndex, LiteLLM, Aider, Claude Code and the OpenAI and Anthropic SDKs, set up step by step.
Find your toolQuestions.
What is uncensored AI?
Uncensored AI is a language model with its refusal behaviour taken out, so it answers the request instead of declining or lecturing. On Wild West API that means two kinds of model: abliterated ones, whose refusal direction was removed from the weights, and uncensored builds, whose refusal training was taken off. The uncensored AI guide covers how it works and which models to use.
What if a model refuses my request?
Email support@wildwestapi.com with the request, either its time or the request id from your dashboard, and we will refund what that request cost to your balance.
What are uncensored AI models?
Open-weight language models with the refusal behaviour taken out, so they answer instead of declining. Some are abliterated, meaning the refusal direction was removed from the weights directly. The others are uncensored builds whose refusal training was taken off. Every model on Wild West API is one or the other. More in abliterated models.
The terms still apply: nothing illegal, nothing sexual involving minors, and nothing aimed at real people.
Does it work with SillyTavern and my other tools?
Yes, anything that takes an OpenAI-compatible base URL: SillyTavern, the OpenAI SDKs, Cline, opencode, Open WebUI and most others. Claude Code and the Anthropic SDKs use the Anthropic-compatible endpoint with the same key. See the setup guides.
What about apps that run in the browser?
Some sites call an API straight from your browser tab, and browsers only allow that when the API answers cross-origin requests. Wild West API does not do that today, so those sites show a network or CORS error. Call it from a server or a local app such as SillyTavern instead.
How do I pay?
Top up prepaid credit from the billing page, with crypto or card. The smallest top-up is $5.00. There is no subscription. Pricing has the details.