UncensoredLabs/uncensored-ai-api

Uncensored AI API — UncensoredLabs

Access uncensored and abliterated DeepSeek and GLM text models through one OpenAI-compatible API, or use them directly in browser Chat. Keep your client, change the base URL, and choose a model.

Get started · Models and pricing · Browser Chat · Documentation

Built in Switzerland. This is the official public repository for guides, configuration examples and API samples. The hosted platform's implementation is not included.

Start with a working request

Create an account and an application API key in the dashboard. A 16-digit account sign-in key is a different credential; do not use it in API requests.

export UNCENSOREDLABS_API_KEY="YOUR_APPLICATION_API_KEY"
curl --fail-with-body https://uncensoredlabs.dev/v1/chat/completions \
  -H "Authorization: Bearer $UNCENSOREDLABS_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "model": "deepseek-v4.1-flash-1017-abliterated",
    "messages": [{"role": "user", "content": "Review this security design: customer comments are rendered as HTML in an admin console. Explain the attack path and how to verify the fix."}],
    "max_tokens": 1024
  }'

Eligible new accounts receive $0.10 in trial credits; repeat-trial signals can require identity verification. Prepaid top-ups start at $3. Check the dashboard's available balance: requests reserve credit before generation, and a small balance may not cover every model's reservation.

Models

Snapshot checked November 1, 2026 against the public API catalog. USD per one million tokens, before any active plan discount. Context is 1,000,000 tokens for every model below; input and output must fit the provider's context accounting.

| Exact API model ID | Input | Cached input | Output | Maximum output | | --- | ---: | ---: | ---: | ---: | | deepseek-v4-pro-uncensored | $0.50 | $0.10 | $0.50 | 65,536 | | deepseek-v4.1-flash-1017-abliterated | $0.30 | $0.12 | $0.30 | 65,536 | | deepseek-v4.1-flash-abliterated | $0.30 | $0.10 | $0.30 | 65,536 | | deepseek-v4.1-flash-abliterated-fast | $1.40 | $0.32 | $1.40 | 65,536 | | glm-5.3-abliterated | $1.40 | $0.40 | $1.40 | 130,000 | | glm-5.3-uncensored | $0.35 | $0.10 | $0.35 | 65,536 |

All six catalog entries support text, streaming and function tools. Catalog inclusion does not mean a model is currently healthy. See model details and performance for live links and measurement limitations. Rates and availability can change; the live catalog takes precedence.

What to build

• Security research: reason through vulnerability chains, review application trust boundaries, and turn findings into regression tests.
• Reverse engineering: explain decompiled routines, recover file-format structures, and compare a vulnerable function with its patch.
• Developer tools: connect a coding assistant, process tool calls, or integrate streamed text into your own interface.
• Creative work: explore difficult themes and unconventional characters with models designed for fewer refusals.
• Trust and safety: create labeled evaluation cases and analyze moderation failure modes.

“Abliterated” describes models whose weights have been modified to reduce refusal behavior. It does not promise a correct answer or no refusals on every prompt. UncensoredLabs is an independent service, not affiliated with OpenAI, DeepSeek or Zhipu AI.

Guides and examples

| Guide | What you get | | --- | --- | | Quickstart | Linux/macOS and PowerShell environment setup; first request | | Python | OpenAI SDK request with automatic retries disabled | | TypeScript | OpenAI SDK request with typed error handling | | Streaming | Incremental text, usage events and stream failures | | Streaming Python | Runnable streamed response | | Streaming TypeScript | Runnable streamed response | | Coding clients | Cline, Continue and OpenCode configuration | | Model catalog | Limits, pricing, live performance and model page links | | Security research workflows | Concrete prompts for analysis and evaluation | | Errors and billing | HTTP errors, rate limits, reservations and retry decisions | | Compatibility and privacy | Supported protocol and content handling |

Stay in touch

X / @uncensoredlabs0 · Telegram bot · Website

Use issues for reproducible documentation problems and example improvements. Keep account, payment and vulnerability details out of public issues; see support and security reporting.

License

The original documentation and example code in this repository are MIT licensed. That license does not cover hosted platform source, model weights, provider services or third-party trademarks.