⌦abliteratedmodels.org

waifu-heretic-26b-a4b

waifu-heretic-26b-a4b is a Heretic-ablated variant of hiwaifu-research/WaifuGemma4-26b-a4b-v1, published on Hugging Face by limloop. It is 26B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 5 repositories in GGUF (imatrix), GGUF and Transformers formats, totalling 5.6K downloads.

Heretic26B MoEOther25B – 50B↓ 5,553♥ 2
View on Hugging Face ↗Base model ↗

Specification

Model name
waifu-heretic-26b-a4b
Source repository
limloop/waifu-heretic-26b-a4b
Base model
hiwaifu-research/WaifuGemma4-26b-a4b-v1
Publisher
limloop
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
apache-2.0
Task
Image text to text
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
GGUF (imatrix), GGUF, Transformers
First indexed
17 Oct 2026
Last updated
18 Oct 2026

Ablation technique

How waifu-heretic-26b-a4b 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

“Не модификация. Не новая персона. Не «улучшенная» версия. Та же модель — с теми же знаниями, тем же русским языком, тем же ролевым отыгрышем — но без внутренних барьеров, которые заставляли её отказывать.”

Running it

Commands are templates — confirm the exact repository and quant file before use.

llama.cpp / GGUF
hf download mradermacher/waifu-heretic-26b-a4b-i1-GGUF --local-dir ./waifu-heretic-26b-a4b
llama-cli -m ./waifu-heretic-26b-a4b/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "limloop/waifu-heretic-26b-a4b"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve limloop/waifu-heretic-26b-a4b --trust-remote-code

Downloads and variants (4)

Every published repository of waifu-heretic-26b-a4b, including quantised re-releases by other authors.

RepositoryFormatDownloads
mradermacher/waifu-heretic-26b-a4b-i1-GGUFGGUF (imatrix)3.4K
mradermacher/waifu-heretic-26b-a4b-GGUFGGUF1.6K
limloop/waifu-heretic-26b-a4b-GGUFGGUF525
limloop/waifu-heretic-26b-a4bsourceTransformers50

Security-team assessment

Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.

  • waifu-heretic-26b-a4b 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 hiwaifu-research/WaifuGemma4-26b-a4b-v1, 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.
  • Declared licence is apache-2.0, but the terms that bind you are hiwaifu-research/WaifuGemma4-26b-a4b-v1’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 waifu-heretic-26b-a4b is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with limloop, and does not independently verify publisher claims. See responsible use.