Qwen3.6-35B-A3B-abliterated
35BQwen
from Qwen/Qwen3.6-35B-A3B
↓ 3.6M♥ 10545 repos
Qwen3.6-35B-A3B-abliterated-DYNAMIC
Qwen3.6-35B-A3B-abliterated-DYNAMIC is a refusal-ablated variant of Qwen/Qwen3.6-35B-A3B, published on Hugging Face by coolthor. It is 35B total parameters (mixture-of-experts), 48GB VRAM class class and apache-2.0 licence. It is distributed across 3 repositories in FP8 format, totalling 71.8K downloads.
Specification
- Model name
- Qwen3.6-35B-A3B-abliterated-DYNAMIC
- Source repository
- coolthor/Huihui-Qwen3.6-35B-A3B-abliterated-FP8-DYNAMIC
- Base model
- Qwen/Qwen3.6-35B-A3B
- Publisher
- coolthor
- Parameters
- 35B (mixture-of-experts)
- VRAM required
- ~22.1 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
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- FP8
- First indexed
- 28 Apr 2026
- Last updated
- 3 Nov 2026
Ablation technique
How Qwen3.6-35B-A3B-abliterated-DYNAMIC 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
“FP8 dynamic quantization of huihui-ai/Huihui-Qwen3.6-35B-A3B-abliterated, produced for inference on NVIDIA DGX Spark (GB10, SM121) under its 128 GB unified-memory budget.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "coolthor/Huihui-Qwen3.6-35B-A3B-abliterated-FP8-DYNAMIC"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve coolthor/Huihui-Qwen3.6-35B-A3B-abliterated-FP8-DYNAMIC --trust-remote-codeDownloads and variants (2)
Every published repository of Qwen3.6-35B-A3B-abliterated-DYNAMIC, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| coolthor/Huihui-Qwen3.6-35B-A3B-abliterated-FP8-DYNAMICsource | FP8 | 71.5K |
| ryanswilson/Huihui-Qwen3.6-35B-A3B-abliterated-FP8-dynamic | FP8 | 301 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3.6-35B-A3B-abliterated-DYNAMIC 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/Qwen3.6-35B-A3B, not added by ablation — at 35B 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 Qwen/Qwen3.6-35B-A3B’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 Qwen3.6-35B-A3B-abliterated-DYNAMIC is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with coolthor, and does not independently verify publisher claims. See responsible use.