⌦abliteratedmodels.org

Gemory-26B-A4B-SOMPOA

Gemory-26B-A4B-SOMPOA is a Heretic-ablated variant of 26B-Suite/Gemory-26B-A4B-SOMPOA, published on Hugging Face by 26B-Suite. It is 26B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 4 repositories in Transformers, GGUF and GGUF (imatrix) formats, totalling 7.3K downloads.

Heretic26B MoEOther25B – 50B↓ 7,299♥ 5
View on Hugging Face ↗Base model ↗

Specification

Model name
Gemory-26B-A4B-SOMPOA
Source repository
26B-Suite/Gemory-26B-A4B-SOMPOA
Base model
26B-Suite/Gemory-26B-A4B-SOMPOA
Publisher
26B-Suite
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
Text generation
Context window
Not documented
Ablation scope
Not stated by the publisher
Formats available
Transformers, GGUF, GGUF (imatrix)
First indexed
28 Oct 2026
Last updated
29 Oct 2026

Ablation technique

How Gemory-26B-A4B-SOMPOA 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

“A creative-writing and roleplay operative built to enter the scene, read the room, make the move, and keep going. Gemory carries 26B total parameters with roughly 4B active per token—a little femme fatale with very little interest in wasting motion.”

Running it

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

llama.cpp / GGUF
hf download mradermacher/Gemory-26B-A4B-SOMPOA-i1-GGUF --local-dir ./gemory-26b-a4b-sompoa
llama-cli -m ./gemory-26b-a4b-sompoa/<file>.gguf -p "..."
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer

repo = "26B-Suite/Gemory-26B-A4B-SOMPOA"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
    repo, torch_dtype="auto", device_map="auto"
)
vLLM
vllm serve 26B-Suite/Gemory-26B-A4B-SOMPOA --trust-remote-code

Downloads and variants (3)

Every published repository of Gemory-26B-A4B-SOMPOA, including quantised re-releases by other authors.

RepositoryFormatDownloads
mradermacher/Gemory-26B-A4B-SOMPOA-i1-GGUFGGUF (imatrix)5.8K
mradermacher/Gemory-26B-A4B-SOMPOA-GGUFGGUF1.3K
26B-Suite/Gemory-26B-A4B-SOMPOAsourceTransformers246

Security-team assessment

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

  • Gemory-26B-A4B-SOMPOA 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 26B-Suite/Gemory-26B-A4B-SOMPOA, 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 26B-Suite/Gemory-26B-A4B-SOMPOA’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 Gemory-26B-A4B-SOMPOA is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with 26B-Suite, and does not independently verify publisher claims. See responsible use.