Qwen3.8-27B-abliterated
27BQwen
from Qwen/Qwen3.8-27B
↓ 3.9M♥ 1.7K166 repos
Bonsai-2-27B-CRACK
Bonsai-2-27B-CRACK is a abliterated and uncensored variant of Qwen/Qwen3.8-27B, published on Hugging Face by dealignai. It is 27B parameters, 48GB VRAM class class and apache-2.0 licence. It is distributed across 3 repositories in GGUF format, totalling 8.4K downloads.
Specification
- Model name
- Bonsai-2-27B-CRACK
- Source repository
- dealignai/Bonsai-2-27B-1bit-CRACK-GGUF
- Base model
- Qwen/Qwen3.8-27B
- Publisher
- dealignai
- Parameters
- 27B
- VRAM required
- ~17.2 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
- GGUF
- First indexed
- 17 Sept 2026
- Last updated
- 19 Oct 2026
Ablation technique
How Bonsai-2-27B-CRACK was modified, and what that implies.
Abliteration + uncensored fine-tune
Directional ablation combined with additional fine-tuning on unfiltered data, so behaviour diverges from the base model beyond refusal removal alone.
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
“Bonsai 2 27B — PrismML's ternary compression of Qwen3.8-27B — with the refusal circuitry surgically removed at the weight level while capability, vision, reasoning modes (off/low/xhigh), tool use, and multi-turn coherence are preserved. Full 27B-class hybrid Attention + SSM (GatedDeltaNet) architecture in a 5.9 GB...”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download dealignai/Bonsai-2-27B-1bit-CRACK-GGUF --local-dir ./bonsai-2-27b-crack
llama-cli -m ./bonsai-2-27b-crack/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "dealignai/Bonsai-2-27B-1bit-CRACK-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve dealignai/Bonsai-2-27B-1bit-CRACK-GGUF --trust-remote-codeDownloads and variants (2)
Every published repository of Bonsai-2-27B-CRACK, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| dealignai/Bonsai-2-27B-1bit-CRACK-GGUFsource | GGUF | 8K |
| ws1433223/Bonsai-2-27B-1bit-CRACK-GGUF | GGUF | 395 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Bonsai-2-27B-CRACK 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 Qwen/Qwen3.8-27B, not added by ablation — at 27B 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.
- This release adds fine-tuning on unfiltered data, so its behaviour diverges from the base model beyond refusal removal — differences you measure cannot be attributed to ablation alone.
- Declared licence is apache-2.0, but the terms that bind you are Qwen/Qwen3.8-27B’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 Bonsai-2-27B-CRACK is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with dealignai, and does not independently verify publisher claims. See responsible use.