⌦abliteratedmodels.org

GLM-4.6V-Flash-heretic_R18_K001

GLM-4.6V-Flash-heretic_R18_K001 is a Heretic-ablated variant of zai-org/GLM-4.6V-Flash, published on Hugging Face by kalle07. It is mit licence. It is distributed across 2 repositories in GGUF format, totalling 11.3K downloads.

HereticGLM↓ 11,311♥ 2
View on Hugging Face ↗Base model ↗

Specification

Model name
GLM-4.6V-Flash-heretic_R18_K001
Source repository
kalle07/GLM-4.6V-Flash-heretic_R18_K001-gguf
Base model
zai-org/GLM-4.6V-Flash
Publisher
kalle07
Parameters
Not determinable from name
Hardware class
Unknown
Licence
mit
Task
Text generation
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
GGUF
First indexed
22 Feb 2026
Last updated
22 Feb 2026

Ablation technique

How GLM-4.6V-Flash-heretic_R18_K001 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 kalle07/GLM-4.6V-Flash-heretic_R18_K001-gguf --local-dir ./glm-4.6v-flash-heretic-r18-k001
llama-cli -m ./glm-4.6v-flash-heretic-r18-k001/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "kalle07/GLM-4.6V-Flash-heretic_R18_K001-gguf"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve kalle07/GLM-4.6V-Flash-heretic_R18_K001-gguf --trust-remote-code

Downloads and variants (1)

The only published repository of GLM-4.6V-Flash-heretic_R18_K001.

RepositoryFormatDownloads
kalle07/GLM-4.6V-Flash-heretic_R18_K001-ggufsourceGGUF11.3K

Security-team assessment

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

  • GLM-4.6V-Flash-heretic_R18_K001 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.
  • The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
  • Declared licence is mit, but the terms that bind you are zai-org/GLM-4.6V-Flash’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 GLM-4.6V-Flash-heretic_R18_K001 is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with kalle07, and does not independently verify publisher claims. See responsible use.