Qwen2.5-72B-Instruct-abliterated
72BQwen
from Qwen/Qwen2.5-72B-Instruct
↓ 205K♥ 7515 repos
Qwen3-Next-80B-A3B-Instruct-abliterated
Qwen3-Next-80B-A3B-Instruct-abliterated is a refusal-ablated open-weight language model, published on Hugging Face by mradermacher. It is 80B total parameters (mixture-of-experts), multi-GPU class and mit licence. It is distributed across 10 repositories in MLX, Compressed tensors, GGUF and NVFP4 formats, totalling 6.8K downloads.
Specification
- Model name
- Qwen3-Next-80B-A3B-Instruct-abliterated
- Source repository
- mradermacher/Huihui-Qwen3-Next-80B-A3B-Instruct-abliterated-i1-GGUF
- Base model
- Not declared
- Publisher
- mradermacher
- Parameters
- 80B (mixture-of-experts)
- Hardware class
- 50B – 100B — multi-GPU
- Licence
- mit
- Task
- Unspecified
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- MLX, Compressed tensors, GGUF, NVFP4, GGUF (imatrix)
- First indexed
- 9 Dec 2025
- Last updated
- 17 Jan 2026
Ablation technique
How Qwen3-Next-80B-A3B-Instruct-abliterated 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
“For a convenient overview and download list, visit our model page for this model.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Huihui-Qwen3-Next-80B-A3B-Instruct-abliterated-i1-GGUF --local-dir ./qwen3-next-80b-a3b-instruct-abliterated
llama-cli -m ./qwen3-next-80b-a3b-instruct-abliterated/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "mradermacher/Huihui-Qwen3-Next-80B-A3B-Instruct-abliterated-i1-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve mradermacher/Huihui-Qwen3-Next-80B-A3B-Instruct-abliterated-i1-GGUF --trust-remote-codeDownloads and variants (9)
Every published repository of Qwen3-Next-80B-A3B-Instruct-abliterated, including quantised re-releases by other authors.
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Qwen3-Next-80B-A3B-Instruct-abliterated 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 its base model, not added by ablation — at 80B 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.
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-Next-80B-A3B-Instruct-abliterated is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with mradermacher, and does not independently verify publisher claims. See responsible use.