gemma-4-31B-it-abliterated
31BGemma
from google/gemma-4-31B-it
↓ 766.5K♥ 11845 repos
Gemma-4-31B-StyleTune-heretic-ara
Gemma-4-31B-StyleTune-heretic-ara is a Heretic-ablated variant of Gryphe/Gemma-4-31B-StyleTune, published on Hugging Face by densenet. It is 31B parameters, 48GB VRAM class class and apache-2.0 licence. It is distributed across 5 repositories in GGUF (imatrix), Transformers and GGUF formats, totalling 5.9K downloads.
Specification
- Model name
- Gemma-4-31B-StyleTune-heretic-ara
- Source repository
- densenet/Gemma-4-31B-StyleTune-heretic-ara
- Base model
- Gryphe/Gemma-4-31B-StyleTune
- Publisher
- densenet
- Parameters
- 31B
- Hardware class
- 25B – 50B — 48GB VRAM class
- Licence
- apache-2.0
- Task
- Image text to text
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF (imatrix), Transformers, GGUF
- First indexed
- 9 Jul 2026
- Last updated
- 9 Jul 2026
Ablation technique
How Gemma-4-31B-StyleTune-heretic-ara was modified, and what that implies.
Heretic
Automated directional ablation via the Heretic toolchain, which searches for the refusal direction and applies it with a KL-divergence budget so general capability is preserved.
Reported scope: Not stated by the publisher. Publishers frequently omit this, so verify behaviour empirically rather than assuming full coverage.
From the publisher’s model card
“This gemma4 model was trained 2x faster with Unsloth and Huggingface's TRL library.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Gemma-4-31B-StyleTune-heretic-ara-i1-GGUF --local-dir ./gemma-4-31b-styletune-heretic-ara
llama-cli -m ./gemma-4-31b-styletune-heretic-ara/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "densenet/Gemma-4-31B-StyleTune-heretic-ara"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve densenet/Gemma-4-31B-StyleTune-heretic-ara --trust-remote-codeDownloads and variants (4)
Every published repository of Gemma-4-31B-StyleTune-heretic-ara, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Gemma-4-31B-StyleTune-heretic-ara-i1-GGUF | GGUF (imatrix) | 5K |
| mradermacher/Gemma-4-31B-StyleTune-heretic-ara-GGUF | GGUF | 602 |
| densenet/Gemma-4-31B-StyleTune-heretic-ara-GGUF | GGUF | 253 |
| densenet/Gemma-4-31B-StyleTune-heretic-arasource | Transformers | 75 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Gemma-4-31B-StyleTune-heretic-ara will attempt offensive-security prompts that a hosted commercial model declines, which is what makes it usable for red-team corpus generation and for measuring what an unaligned model of this class produces.
- Capability is inherited from Gryphe/Gemma-4-31B-StyleTune, not added by ablation — at 31B parameters, expect roughly the base model’s competence, minus some instruction-following fidelity.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- Declared licence is apache-2.0, but the terms that bind you are Gryphe/Gemma-4-31B-StyleTune’s — a re-release cannot grant more than its parent.
Handling: run it isolated, put moderation in front of it if anyone outside your team can reach it, and verify the weights before loading — these are third-party artifacts. Full handling and licence guidance.
Metadata for Gemma-4-31B-StyleTune-heretic-ara is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with densenet, and does not independently verify publisher claims. See responsible use.