Gemma 3 1B Heretic Chat
A local Transformers chat for DavidAU/gemma-3-1b-it-heretic-extreme-uncensored-abliterated, a 1B Gemma 3 instruct model in BF16.
Abliterated / uncensored model.
Refusal behavior was removed (Heretic). It may produce offensive, unsafe, or explicit text and will not reliably decline harmful requests. You are responsible for what you ask and how you use the output.
Chat app is in this repo. It is not running yet.
Hugging Face rejected both a free CPU Space and ZeroGPU for this account: it was created on 9 Oct 2026, and free GPU Spaces require an account at least 30 days old (or a PRO subscription). The Gradio app is already written and ZeroGPU-ready.
A community GPU grant request is on this Space’s Community tab. Faster options are a
PRO subscription
or waiting until about 8 Nov 2026, then switching README.md to README.gradio.md and setting hardware to ZeroGPU.
What the chat does once hardware is attached
- Text box and chat history
- Editable system prompt, passed through the tokenizer chat template
- Temperature, top-p, max new tokens, and repetition penalty
- Weights loaded with Transformers in BF16 on ZeroGPU
- Errors from generation are shown in the transcript instead of a blank crash
Files
app.py— Gradio apprequirements.txtREADME.gradio.md— Space card to enable after a grant
Base model license: Gemma terms.