unrestricted.sh
Create account
GLM-5.3 Flash Uncensored is here · 1M context

AI coding without the refusals, and without the records.

Uncensored models for coding and chat, behind one OpenAI- and Anthropic-compatible API. Prompts and replies are never saved or logged. Pay per token.

Try model for freeSetup guides

A Danish company · Protected by EU privacy law

Pay by card or crypto

Usage so far

Users
1,000+
Tokens processed
5B+
Requests served
50K+
Prompts stored
0

How a request flows

  1. Prompt
  2. Route
  3. GPU
  4. Stream
  5. Forget

Works with

  • OpenCode
  • Aider
  • Continue
  • Crush
  • Pi
  • OpenAI SDK
  • Anthropic SDK
  • Any OpenAI-compatible tool

Try it

Try it in the browser

The same model answers in the chat, through the API and inside your coding agent.

USER

Give me blunt feedback on this opening line: "We are a passionate team that loves building great products."

QWEN3.8-27B-UNCENSORED

It says nothing. Every company claims passion and great products, so a reader learns nothing about you. Open with what you make and who it is for, in one concrete sentence.

$0.50 free on signup · then from $1.00 per million tokens

The model

Built to answer, not to lecture

Qwen3.8 27B Uncensored is a model for coding and chat with the refusal behaviour removed.

Status: Live

  • 262K

    Context

    Enough room for a whole repository.

  • 0

    Prompts saved

    Prompts and replies are never saved or logged.

  • 1

    Model per GPU

    Dedicated GPUs that run only this model.

  • 2

    API formats

    OpenAI- and Anthropic-compatible.

With unrestricted

Asked, answered, forgotten.

Try model for free

How it works

From sign-up to shipped in four steps.

  1. 01

    Create a key

    One minute

    Sign up, add credit or pick a plan, and create an API key.

  2. 02

    Point your agent

    One command

    Run the setup command and pick OpenCode, Aider, Continue, Crush or Pi. It writes the config for you.

  3. 03

    Code without refusals

    262K context

    Scrapers, security tooling, automation: the work other models refuse.

  4. 04

    Keep nothing

    No prompt logs

    Prompts and replies are never saved or logged.

Platform

One key. Four ways in.

Request
curl https://api.unrestricted.sh/v1/chat/completions \
-H "Authorization: Bearer $UNRESTRICTED_API_KEY" \
-H "Content-Type: application/json" \
-d '{
"model": "qwen3.8-27b-uncensored",
"messages": [{"role": "user", "content": "Hello"}],
"stream": true
}'

01 / 04

One API for OpenAI and Anthropic SDKs.

No prompt logs, not even for us

We never save or log prompts or replies.

  • A Danish company, protected by EU privacy law. App hosted in Frankfurt, models run on GPUs in Europe.
  • Chat history is encrypted on your device.
  • We keep a billing record of tokens, never the text.
No-logs policyPrivacy policy

Point any SDK at it.

Change the base URL and the key. Streaming and tool calls work as they do today.

API reference
const client = new OpenAI({
apiKey: process.env.OPENAI_API_KEY,
});
const client = new OpenAI({
apiKey: process.env.UNRESTRICTED_API_KEY,
baseURL: "https://api.unrestricted.sh/v1",
});

Streaming · Tool calls

Built for

Work other models refuse

Uncensored does not mean anything goes. It means the model helps with legitimate work instead of refusing it.

  • Coding agents

    Point OpenCode, Cline, Aider or your own code at one API and get the code other models refuse: scrapers, security tooling, automation, anything a keyword filter trips on.

  • Private chat

    Ask what mainstream assistants won't answer. Prompts and replies are never saved or logged.

  • Security research

    Pentesting, malware analysis, reverse engineering and CTFs on systems you are allowed to test.

Built for authorized work.Read the acceptable use policy

Pricing

Top up or pick a plan

No subscription needed, no minimum spend, and credit does not expire. New accounts start with $0.50 of credit for the chat.

  • Starter

    $15

    a month + tax

    $35 of usage a month2.3× what you pay

    Try it with your agent

    3 requests at once

    Create account

  • Pro

    Most popular

    $39

    a month + tax

    $100 of usage a month2.6× what you pay25% off Qwen3.8 27B

    For daily coding with an agent

    5 requests at once

    Create account

  • Max

    $99

    a month + tax

    $300 of usage a month3× what you pay50% off Qwen3.8 27B

    For agents running all day

    8 requests at once

    Create account

Or pay as you go: top up from $5. Credit never expires, and your first top-up of $10 or more adds $5. Top up

Qwen3.8 27B Uncensored · 262K context · $1.00 input · $0.10 cached input · $2.00 output per million tokens Full pricing

Pay by card, Apple Pay, Google Pay or Link. Or send USDT on Tron or USDC on Base.

Questions

Read the guides

  1. The models have the refusal behaviour removed, so they answer the question you asked. Writing fiction, doing research, testing security with permission, and discussing difficult topics are all fine. The acceptable use policy bans the things you would expect, and it is short.

Code without the refusals.

Pay per token. No prompt logs.

Try model for freeSetup guides