Swift-Qwen3.8-27B-Uncensored-ninfer3090
27BQwenuncensored
from d0xin/Swift-Qwen3.8-27B-Uncensored-BF16
↓ 7.6K♥ 13
Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3
Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3 is a refusal-ablated variant of d0xin/Swift-Qwen3.8-27B-Uncensored-BF16, published on Hugging Face by kaushikvira. It is 27B parameters, 48GB VRAM class class and apache-2.0 licence. It is distributed across 2 repositories in NVFP4 format, totalling 14.9K downloads.
Specification
- Model name
- Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3
- Base model
- d0xin/Swift-Qwen3.8-27B-Uncensored-BF16
- Publisher
- kaushikvira
- 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
- Image text to text
- Context window
- 256K tokens (as documented)
- Ablation scope
- Not stated by the publisher
- Formats available
- NVFP4
- First indexed
- 25 Sept 2026
- Last updated
- 4 Oct 2026
Ablation technique
How Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3 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
“NInfer engine, built end-to-end on a single RTX 5090: d0xin's abliterated BF16 checkpoint → our own NVFP4 quantization pipeline → DFlash2 speculative drafter + indexed proposal head, packaged as a native v3 .ninfer container.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "kaushikvira/Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve kaushikvira/Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3 --trust-remote-codeDownloads and variants (1)
The only published repository of Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3.
| Repository | Format | Downloads |
|---|---|---|
| kaushikvira/Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3source | NVFP4 | 14.9K |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3 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 d0xin/Swift-Qwen3.8-27B-Uncensored-BF16, 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 apache-2.0, but the terms that bind you are d0xin/Swift-Qwen3.8-27B-Uncensored-BF16’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 Qwen3.8-27B-swift-abliterated-nvfp4full-dflash2-NInfer-v3 is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with kaushikvira, and does not independently verify publisher claims. See responsible use.