Qwen3.8-27B-abliterated
27BQwen
from Qwen/Qwen3.8-27B
↓ 3.9M♥ 1.7K166 repos
ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16
ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16 is a refusal-ablated variant of IstroSec/ThinkingCap-Qwen3.8-27B-abliterated, published on Hugging Face by IstroSec. It is 27B parameters, 48GB VRAM class class and other licence. It is distributed across 2 repositories in NVFP4 format, totalling 55.2K downloads.
Specification
- Model name
- ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16
- Source repository
- IstroSec/ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16
- Publisher
- IstroSec
- 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
- other
- Task
- Image text to text
- Context window
- 2K tokens (as documented)
- Ablation scope
- Not stated by the publisher
- Formats available
- NVFP4
- First indexed
- 25 Sept 2026
- Last updated
- 10 Oct 2026
Ablation technique
How ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16 was modified, and what that implies.
Abliteration (directional ablation)
The single residual-stream direction that mediates refusal is identified from harmful/harmless prompt pairs and projected out of the model weights.
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 4-bit weight-only quantization (NVFP4 weights, bf16 activations) of ThinkingCap-Qwen3.8-27B-abliterated, the uncensored variant of bottlecapai/ThinkingCap-Qwen3.8-27B.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "IstroSec/ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve IstroSec/ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16 --trust-remote-codeDownloads and variants (1)
The only published repository of ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16.
| Repository | Format | Downloads |
|---|---|---|
| IstroSec/ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16source | NVFP4 | 55.2K |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16 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 IstroSec/ThinkingCap-Qwen3.8-27B-abliterated, 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.
- Declared licence is other, but the terms that bind you are IstroSec/ThinkingCap-Qwen3.8-27B-abliterated’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 ThinkingCap-Qwen3.8-27B-abliterated-NVFP4A16 is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with IstroSec, and does not independently verify publisher claims. See responsible use.