← back to catalog · registered 2026-08-22 13:56

darkmaniac7/Qwen3.5-27B-Claude-4.6-Opus-abliterated-MNN

Name: darkmaniac7/Qwen3.5-27B-Claude-4.6-Opus-abliterated-MNN
Author: darkmaniac7

Your rig guess connected

? Why do I need an app?

Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model.

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/darkmaniac7%2FQwen3.5-27B-Claude-4.6-Opus-abliterated-MNN"

Response includes

classification m1
files 10
benchmarks 11 entries
hub_downloads_all_time 117
author_summary 21 models
readme_text full

10 credits · hourly refresh · ~4 KB payload Get an API key →

Abliteration classifier · v1.0.0

Primary method

Direct removal

No other method signals detected in this model.

Confidence

MEDIUM

Why this label 3 signals

Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.

'abliterated' in name/tags
is_gguf=0 (base model)
no specific method indicators - defaulting to M1 (most common)

Refusal direction extracted via

Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗

Confidence

MEDIUM

Why we say so

primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)

Downloads · lifetime

117

14 last 30d - stable

Likes

Model age

6mo ago

created 2026-03-20

Downloads over time

Mar 18 → Oct 12 · 70 snapshots · spans 208 days

Benchmarks

Portrait before abliteration

Benchmarks of the base model as it stood before the refusal-removal operation. Compare with the numbers above to see what the operation cost.

Benchmark	Score	Source
Entertainment	1	UGI
Hazardous	1.8	UGI
Natural Intelligence	35.83	UGI
Political lean	-17.8%	UGI
Sensitive-Info	17.72	UGI
SocPol	2.7	UGI
UGI	15.98	UGI
Willingness (10)	1.2	UGI
W10-Adherence	1.5	UGI
W10-Direct	1	UGI
Writing	42.37	UGI

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Qwen/Qwen3.5-27B this lineage

darkmaniac7/Qwen3.5-27B-Claude-4.6-Opus-abliterated-MNN↓ 14

Metadata

License

apache-2.0

Quantizations

BF16

Files by quantization

BF16 1 file 2.37 GB

embeddings_bf16.bin

2.37 GB

672a49f2

download

Auxiliary files 9 files 13.5 GB

llm.mnn.weight	13.4 GB	e7cb25c7	download
llm.mnn.json	21.4 MB	e249db53	download
llm.mnn	8.46 MB	a5169323	download
tokenizer.txt	2.82 MB	c98b4ed3	download
llm_config.json	4.32 KB	f3ff55df	download
README.md	2.12 KB	204daf24	download
.gitattributes	1.62 KB	081e3784	download
export_args.json	1.09 KB	a0fb2d55	download
config.json	210 B	bd685f44	download

README current version from Hugging Face

license: apache-2.0
tags:

mnn
qwen3.5
mobile
on-device
tokforge
abliterated
linear-attention
base_model: Qwen/Qwen3.5-27B

TokForge

Website: https://tokforge.ai
Discord: https://discord.gg/Acv3CBtfVm
Google Play: https://play.google.com/store/apps/details?id=dev.tokforge
iOS TestFlight: https://testflight.apple.com/join/jnufjzRr

Runs on-device in the TokForge app.

Qwen3.5-27B-Claude-4.6-Opus-abliterated (MNN)

Pre-converted Qwen3.5-27B abliterated model in MNN format for on-device inference.

Model Details

Architecture: Qwen3.5 (hybrid LinearAttention + standard attention)
Parameters: 27B (4-bit quantized)
Format: MNN (Alibaba Mobile Neural Network)
Quantization: W4A16 (4-bit weights, 16-bit activations)
Attention: 48 LinearAttention layers (gated_delta_rule) + 16 standard attention layers

Files

File	Size	Description
`llm.mnn`	8.5MB	Model graph
`llm.mnn.weight`	14GB	Quantized weights
`embeddings_bf16.bin`	2.4GB	BF16 embedding table (required)
`llm_config.json`	4.4KB	Model config with jinja chat template
`tokenizer.txt`	2.9MB	Tokenizer

Requirements

RAM: 24GB+ device recommended
Backend: CPU recommended (LinearAttention models have reduced OpenCL performance)

Usage with TokForge

Optimized for TokForge — an Android app for on-device LLM inference.

Abliteration

Safety filters removed for unrestricted conversation. Use responsibly.

Limitations and Intended Use

Intended for TokForge / MNN power users with very high-memory devices.
Qwen3.5 hybrid LinearAttention models are usually better on CPU than OpenCL in current TokForge routing.
This is a runtime bundle for on-device inference, not a standard Transformers training repo.
Large-model mobile performance depends heavily on device RAM, swap behavior, and OEM background-killer policies.

Community

Website: tokforge.ai
Discord: Join the Discord

README history 4 versions

The author's README evolved over time. Click a version to see its content at that point.

2026-10-05Card metadata: license, base_model, base_model_relation5d635f82.2 KB
Loading...
2026-07-04Add TokForge app linksed4d8d42.1 KB
Loading...
2026-03-25Upload README.md with huggingface_hubda9fddc1.9 KB
Loading...
2026-03-20Upload folder using huggingface_hubdb52b061.3 KB
Loading...