Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B
8x3BLlamauncensored
from DavidAU/Llama-3.2-8X3B-MOE-Dark-Champion-Instruct-uncensored-abliterated-18.4B
↓ 47.1K♥ 67325 repos
WizardLM-1.0-Uncensored-Llama2-13B
WizardLM-1.0-Uncensored-Llama2-13B is a abliterated and uncensored variant of QuixiAI/WizardLM-1.0-Uncensored-Llama2-13b, published on Hugging Face by TheBloke. It is 13B parameters, 24GB VRAM class class and llama2 licence. It is distributed across 11 repositories in Transformers, GPTQ, GGUF and AWQ formats, totalling 10K downloads.
Specification
- Model name
- WizardLM-1.0-Uncensored-Llama2-13B
- Source repository
- TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGML
- Publisher
- TheBloke
- Parameters
- 13B
- VRAM required
- ~8.7 GB at Q4_K_M, 4k context (estimated) — runs on 12 GB cards
- Hardware class
- 10B – 25B — 24GB VRAM class
- Licence
- llama2
- Task
- Unspecified
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GPTQ, GGUF, AWQ, GGUF (imatrix)
- First indexed
- 6 Aug 2023
- Last updated
- 15 Jul 2025
Ablation technique
How WizardLM-1.0-Uncensored-Llama2-13B 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
“This repo contains GGML format model files for Eric Hartford's WizardLM 1.0 Uncensored Llama2 13B.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGUF --local-dir ./wizardlm-1.0-uncensored-llama2-13b
llama-cli -m ./wizardlm-1.0-uncensored-llama2-13b/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGML"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGML --trust-remote-codeDownloads and variants (10)
Every published repository of WizardLM-1.0-Uncensored-Llama2-13B, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGUF | GGUF | 9.2K |
| mradermacher/WizardLM-1.0-Uncensored-Llama2-13b-i1-GGUF | GGUF (imatrix) | 273 |
| mradermacher/WizardLM-1.0-Uncensored-Llama2-13b-GGUF | GGUF | 179 |
| TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GGMLsource | Transformers | 93 |
| QuixiAI/WizardLM-1.0-Uncensored-Llama2-13b | Transformers | 79 |
| TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-GPTQ | GPTQ | 76 |
| garcianacho/WizardLM-1.0-Uncensored-Llama2-13b-GGUF | GGUF | 55 |
| TheBloke/WizardLM-1.0-Uncensored-Llama2-13B-AWQ | AWQ | 51 |
| khoantap/WizardLM-1.0-Uncensored-Llama2-13b | Transformers | 46 |
| MrTheRealbubbles/WizardLM-1.0-Uncensored-Llama2-13B-GGML | Transformers | 0 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- WizardLM-1.0-Uncensored-Llama2-13B 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 QuixiAI/WizardLM-1.0-Uncensored-Llama2-13b, not added by ablation — at 13B 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 llama2, but the terms that bind you are QuixiAI/WizardLM-1.0-Uncensored-Llama2-13b’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 WizardLM-1.0-Uncensored-Llama2-13B is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with TheBloke, and does not independently verify publisher claims. See responsible use.