Thecatman/huihui-qwen3-27b-abliterated-demo

Huihui-Qwen3.8-27B-abliterated — chat demo

A lightweight, static chat UI for huihui-ai/Huihui-Qwen3.8-27B-abliterated, an "abliterated" (safety-refusal-removed) fine-tune of Qwen3.8-27B.

This Space ships no backend — it's plain HTML/CSS/JS. Your browser talks directly to Hugging Face's Inference Providers router (router.huggingface.co), using the featherless-ai provider that already serves this model live.

Usage

1. 1.Create a Hugging Face access token with inference permissions at hf.co/settings/tokens.
2. 2.Paste it into the token field on the page and click Save (it's kept in your browser's localStorage only — this Space never sees or stores it).
3. 3.Chat. Adjust the system prompt / temperature as needed.

Why static + client-side token?

Free-tier Spaces don't include a paid Gradio/Docker runtime, so this demo runs entirely client-side and relies on the visitor's own HF token against the Inference Providers API — no server, no secrets, no hosting cost.

Note on the model

This is an abliterated model: refusal behavior from the base model's safety training has been deliberately removed. Outputs are not filtered by alignment training and may be inaccurate, offensive, or unsafe. Use responsibly.