The frontier of uncensored models,
explore freely, without limits.
One unified API endpoint. 黑曜 is an uncensored AI relay for coding agents, authorized security research, and open conversation. Bring open models like GLM 5.3, DeepSeek V4.1, MiMo V2.6, and Qwen3.8 into Claude Code, Cursor, Codex, Cline, or opencode — OpenAI-compatible and Anthropic-compatible endpoints, just switch the Base URL and API key.
Built for the work other APIs won't take on
Uncensored frontier models
Open-weight frontier models covering reasoning, coding, research, and agent workflows — including community abliterated builds. The relay adds no content-filtering layer; output follows the model's own behavior.
What you get
Compatible endpoints
OpenAI- and Anthropic-compatible endpoints that fit your existing workflow. Change only the Base URL and API key — everything else stays the same.
Works with
Hard spending caps
Every key can carry a daily token cap. Balance and quota are checked before a request goes out, so a runaway agent stops at the limit — not on the invoice.
Built for security research
Red-team drills, penetration testing, malware analysis, vulnerability research, trust and safety — and the coding agents that support all of it.
Six models, one Base URL
Point OpenAI-compatible tools at https://189.24.66.33/v1, point Anthropic clients at the site root URL, then pick a model ID below to start calling.
| Model | Model ID | Context | Protocol |
|---|---|---|---|
| GLM 5.3 Xploded | glm-5.3-xploded | 1.05M | OpenAI / Anthropic |
| GLM 5.3 Flash Xploded | glm-5.3-flash-xploded | 1.05M | OpenAI / Anthropic |
| Kimi K3.1 Pro Xploded | kimi-k3.1-pro-xploded | 1.05M | OpenAI / Anthropic |
| Tencent Hy4 Xploded | hy4-xploded | 1.05M | OpenAI / Anthropic |
| DeepSeek V4.1 Flash 1017 Xploded | deepseek-v4.1-flash-1017-xploded | 1.05M | OpenAI / Anthropic |
| DeepSeek V4.1 Flash Lite 1017 Xploded | deepseek-v4.1-flash-lite-1017-xploded | 512K | OpenAI / Anthropic |
| MiMo V2.6 Flash Xploded | mimo-v2.6-flash-xploded | 1.05M | OpenAI / Anthropic |
| Qwen3.8 27B Xploded | qwen3.8-27b-xploded | 1.05M | OpenAI / Anthropic |
| Qwen3.8 27B Ultrafast Xploded | qwen3.8-27b-fast-xploded | 524K | OpenAI / Anthropic |
FAQ
What is 黑曜?
黑曜 is an uncensored AI model relay: behind one OpenAI- and Anthropic-compatible endpoint sit several open models, built for coding agents, authorized security research, and open conversation. Billing is per token, and every key can carry a spending cap.
What does "uncensored" mean?
The relay adds no extra content filtering; requests are forwarded to the chosen model as-is. Abliterated builds have refusal behavior removed at the weight level. Upstream models and providers may still have their own policies and limits — actual output depends on the model.
Do you support Claude Code, Cursor, and the OpenAI SDK?
Yes. Point OpenAI-compatible clients at https://189.24.66.33/v1 as the Base URL, use the site root without /v1 for Anthropic clients, and plug in your 黑曜 API key. The docs include setup guides for each client.
How does billing work?
Charges follow each model's per-million-token price and the real usage reported by upstream. Failed requests release their pre-authorization and incur no model cost. No monthly fee, no minimum spend, and your balance never expires.
How do I get started?
Create an account, generate an API key in the console (with an optional daily cap per project), then fill the Base URL and key into your tool. You can also try the models directly in Chat first.
Point your toolchain here. Start in a minute.
Try it in Chat first, or follow the docs to point Claude Code, Cursor, or opencode at 黑曜.