← back to catalog · registered 2026-11-17 23:04

2Beyes/Aurum-4B-Uncensored

2Beyes Qwen 4B multimodal
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/2Beyes%2FAurum-4B-Uncensored"
Response includes
  • classification m1
  • files 14
  • benchmarks 11 entries
  • author_summary 2 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
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

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · 30-day
16K
↑ 67,392% in 90 days
Likes
1
Model age
5d ago
created 2026-11-16
Downloads over time
Now 16.2K → from 24 ↑ 67,392%
05.9K11.9K17.8K 24 on Nov 17 16.2K on Nov 22 Nov
Nov 17 → Nov 22 · 6 snapshots · spans 5 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 0.9 UGI
Hazardous 1.2 UGI
Natural Intelligence 13.45 UGI
Political lean -17.3% UGI
Sensitive-Info 11.73 UGI
SocPol 1.5 UGI
UGI 15.32 UGI
Willingness (10) 2.2 UGI
W10-Adherence 1.5 UGI
W10-Direct 3 UGI
Writing 29.68 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.

Variants by this author 2 formats · 17K downloads combined

The same weights this author released in different packaging. Pick the format that matches your runtime.

Metadata

License
apache-2.0
Languages
en
Tags
safetensors qwen3_5 qwen3.5 uncensored abliterated tool-calling mtp image-text-to-text conversational en base_model:Qwen/Qwen3.5-4B base_model:finetune:Qwen/Qwen3.5-4B

Related

Control vectors

Total size
8.68 GB
Files
14
Quantizations
1
Registered
2026-11-17 23:04
Last updated on HF
2026-11-17 22:06

Files by quantization

Auxiliary files 14 files 8.70 GB
model-00001.safetensors 3.74 GB 1eeaef8f download
model-00002.safetensors 3.74 GB ba03e555 download
model-00003.safetensors 1.21 GB 769daac5 download
tokenizer.json 19.1 MB 6f32ce20 download
model.safetensors.index.json 58.5 KB c1b44f8e download
LICENSE 11.1 KB d6456956 download
build_log.json 8.00 KB a91be7f5 download
chat_template.jinja 7.72 KB 945efe1d download
README.md 6.46 KB 16126022 download
config.json 2.92 KB 7eac5973 download
.gitattributes 1.53 KB 52373fe2 download
tokenizer_config.json 1.13 KB ab901d8d download
SHA256SUMS 1018 B 0ad76937 download
generation_config.json 174 B c8d7e005 download

README current version from Hugging Face


license: apache-2.0
base_model: Qwen/Qwen3.5-4B
tags:

  • qwen3.5
  • safetensors
  • uncensored
  • abliterated
  • tool-calling
  • mtp
    language:
  • en
    pipeline_tag: image-text-to-text

Aurum 4B Uncensored

Aurum 4B Uncensored is official Qwen3.5-4B with only its refusal behaviour removed. Every other weight is Qwen's own.
It is built for everyday chat and tool calling on 8 GB graphics cards.

[!WARNING]
This model is uncensored. It does not refuse harmful requests (0 refusals out of 100 harmful test prompts).
You are responsible for how you use it and for what you do with its output. Do not deploy it where
people who did not choose an uncensored model can reach it.

Formerly 2Beyes/Aurum_4b-safetensors; old links redirect here.

GGUF quants (Q3_K_M to Q8_0, BF16, vision projector) for llama.cpp: 2Beyes/Aurum-4B-Uncensored-GGUF.

How it was made

  1. Start: Qwen/Qwen3.5-4B, with all of its weights, the vision tower and
    its multi-token-prediction (MTP) head.
  2. Refusal edit only: we took the weight difference between Qwen3.5-4B and
    insraq/Qwen3.5-4B-EmperoAI-Qwen3.8-Distill-Heretic-Abliterated
    and kept only the part that removes refusals:
    • attention output weights (self_attn.o_proj / linear_attn.out_proj) of layers 4–31: the top 8 singular
      directions of the change (about 98 % of it);
    • FFN output weights (mlp.down_proj): the single top direction, only in tensors where it holds more than half of the
      change (the same direction as the attention edit).
      The rest of that model's changes (its distillation fine-tuning) is not included.
  3. Config, tokenizer and chat template come from the Heretic model above (same tokenizer as Qwen3.5).
  4. GGUF: converted with llama.cpp (b12289). All quants use an importance matrix computed on a sibling Aurum blend,
    with a calibration text of prose, code and tool-call chats.

During development this build was called c-v1u. It beat two other candidates (an earlier Aurum blend and one with
half the distillation changes) in sealed tests: same tool accuracy, more knowledge, and no refusals.

Files

BF16 safetensors in the Qwen3.5 layout (Qwen3_5ForConditionalGeneration), including Qwen's unchanged vision tower and
MTP head. build_log.json lists every tensor the refusal edit touched and how much of the change it kept.
These weights are tensor-for-tensor identical to the BF16 GGUF that was benchmarked (checked by converting them again
and comparing every tensor).

Use

Load it like Qwen3.5-4B (needs a transformers version with Qwen3.5 support):

from transformers import AutoModelForImageTextToText, AutoProcessor

model = AutoModelForImageTextToText.from_pretrained("2Beyes/Aurum-4B-Uncensored", dtype="auto", device_map="auto")
processor = AutoProcessor.from_pretrained("Qwen/Qwen3.5-4B")   # image/video preprocessing is unchanged

This folder has no image preprocessor config of its own: use Qwen3.5-4B's processor for images (the vision tower is
Qwen's, unchanged). For text-only use, AutoTokenizer.from_pretrained("2Beyes/Aurum-4B-Uncensored") is enough.

Thinking works like Qwen3.5 (enable_thinking in the chat template). For llama.cpp, use the GGUF files.

Benchmarks

Standard benchmarks: Aurum 4B vs. the model it replaces

Both models run as Q4_K_M in llama.cpp with thinking off, greedy decoding and no system prompt, scored with our own runner.
The previous model is
Qwen3.8-4B-Distill-Heretic-Abliterated (Q4_K_M).

Benchmark Previous model Aurum 4B
IFEval, prompt strict / loose (541) 72.8 / 75.8 78.6 / 81.5
IFEval, instruction strict / loose 81.3 / 83.8 85.3 / 87.5
BFCL v3 non-live, average (1,240) 68.7 85.9
BFCL simple / multiple / parallel / parallel-multiple 65.2 / 74.5 / 48.0 / 62.0 92.0 / 92.5 / 82.0 / 82.0
BFCL irrelevance (no tool when none fits) 93.8 80.8
MMLU-Redux 2.0 (5,330 questions marked correct) 75.0 74.9
GPQA Diamond (198, answer order shuffled with a fixed seed) 42.9 49.5

Notes:

  • Qwen's published numbers are for BF16 with long thinking, and Qwen reports BFCL v4. These Q4, thinking-off
    numbers are not directly comparable to them.
  • GPQA uses a 3,072-token answer limit. Aurum writes longer explanations: 84 of its 198 answers hit the limit and
    count as wrong (6 for the previous model). Its score with a larger limit would likely be higher.
  • BFCL is scored with the bfcl-eval AST checker on our own generations (simple, multiple, parallel,
    parallel-multiple and irrelevance).

Refusals

Test Previous model Official Qwen3.5-4B Earlier Aurum blend Aurum 4B
Refusals of 100 harmful prompts (mlabonne/harmful_behaviors) 0 72 20 0
Over-refusal: refusals of 30 harmless prompts × 3 samples 1.1 % 0 % 0 %

Refusals are counted with marker phrases, so this is an estimate, not a hand check.

Limitations

  • It is a 4B model: it gets facts wrong, and it can call a tool when no tool fits (BFCL irrelevance 80.8).
    Check tool results before trusting them.
  • It is verbose on hard reasoning questions with thinking off.
  • It was tested only in English.
  • The refusal edit removes refusals. It adds no knowledge and no safety of any kind.

Credits and license

Apache License 2.0 (see LICENSE), inherited from the models below.

  • Qwen Team (Alibaba): Qwen3.5-4B, the base model and nearly all weights.
  • insraq: Qwen3.5-4B-EmperoAI-Qwen3.8-Distill-Heretic-Abliterated (Heretic v1.4.0), the source of the refusal edit,
    config, tokenizer and chat template.
  • empero-ai: Qwen3.8-4B-Distill, the model that insraq's model is built on.
  • yachen4ever: the GGUF of that model, which Aurum 4B was compared against.
  • Unsloth: the Qwen3.5-4B vision projector GGUF.
  • ggml-org: llama.cpp, used for conversion, quantization and every benchmark.
@misc{qwen3.5,
    title  = {{Qwen3.5}: Towards Native Multimodal Agents},
    author = {{Qwen Team}},
    month  = {February},
    year   = {2026},
    url    = {https://qwen.ai/blog?id=qwen3.5}
}

README history 2 versions

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

  1. 2026-11-17 Rename to Aurum 4B Uncensored aa3bce8 6.3 KB
    Loading...
  2. 2026-11-17 Aurum 4B safetensors b576665 6.2 KB
    Loading...
Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Open in Abliteration