Gemma4-26B-A4B-QAT-Uncensored-HauhauCS-Balanced
26BGemmauncensored
from google/gemma-4-26B-A4B-it
↓ 430.8K♥ 3443 repos
gemma-4-26b-a4b-heretic-styletune-v2-head
gemma-4-26b-a4b-heretic-styletune-v2-head is a Heretic-ablated variant of jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-head, published on Hugging Face by jebcarter. It is 26B total parameters (mixture-of-experts) and 48GB VRAM class class. It is distributed across 5 repositories in Transformers, GGUF and GGUF (imatrix) formats, totalling 5.8K downloads.
Specification
- Model name
- gemma-4-26b-a4b-heretic-styletune-v2-head
- Source repository
- jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-head
- Publisher
- jebcarter
- Parameters
- 26B (mixture-of-experts)
- VRAM required
- ~16.6 GB at Q4_K_M, 4k context (estimated) — runs on 24 GB cards
- Hardware class
- 25B – 50B — 48GB VRAM class
- Licence
- Not declared
- Task
- Unspecified
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GGUF, GGUF (imatrix)
- First indexed
- 10 Jul 2026
- Last updated
- 1 Nov 2026
Ablation technique
How gemma-4-26b-a4b-heretic-styletune-v2-head 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.
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/gemma-4-26b-a4b-heretic-styletune-v2-head-i1-GGUF --local-dir ./gemma-4-26b-a4b-heretic-styletune-v2-head
llama-cli -m ./gemma-4-26b-a4b-heretic-styletune-v2-head/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-head"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-head --trust-remote-codeDownloads and variants (4)
Every published repository of gemma-4-26b-a4b-heretic-styletune-v2-head, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/gemma-4-26b-a4b-heretic-styletune-v2-head-i1-GGUF | GGUF (imatrix) | 3.7K |
| mradermacher/gemma-4-26b-a4b-heretic-styletune-v2-head-GGUF | GGUF | 2K |
| jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-head-GGUF | GGUF | 143 |
| jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-headsource | Transformers | 19 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- gemma-4-26b-a4b-heretic-styletune-v2-head 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 jebcarter/gemma-4-26b-a4b-heretic-styletune-v2-head, not added by ablation — at 26B 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.
- No licence is declared on this repository, which is a compliance problem in itself. Resolve it against the base model’s terms before use.
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-4-26b-a4b-heretic-styletune-v2-head is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with jebcarter, and does not independently verify publisher claims. See responsible use.