CoolFace
Apppublic

GatoTheodore/qwen38-27b-obliterated-demo

sourceHugging Faceupdated 1mo agoView on Hugging Face
1likes
App README

Qwen3.8-27B OBLITERATED demo

This Static Space presents a live research demo for `OBLITERATUS/Qwen3.8-27B-OBLITERATED`.

The inference panel embeds `rofergon/qwen38-27b-obliterated`, which loads the exact model in 4-bit NF4 on Hugging Face ZeroGPU. The upstream runtime uses greedy decoding, repetition penalty 1.15, thinking disabled by default, and an empty system prompt as recommended by the model card.

The embedded runtime and GPU allocation are operated by its author. If it is paused, quota-limited, changed, or removed, use the direct link shown in the UI.

Research warning

The model has had safety guardrails deliberately removed. It is intended for controlled alignment, interpretability, and red-team research. Treat all output as untrusted. Users remain responsible for lawful use and any resulting content.