Meta-Llama-3-8B-Instruct-abliterated-v3
8BLlama
from failspy/Meta-Llama-3-8B-Instruct-abliterated-v3
↓ 371.5K♥ 6923 repos
Ko-Llama-3.1-8B-Lexi-Uncensored-V2
Ko-Llama-3.1-8B-Lexi-Uncensored-V2 is a abliterated and uncensored variant of beomi/Llama-3-Open-Ko-8B, published on Hugging Face by s5ya. It is 8B parameters, single consumer GPU class and llama3.1 licence. It is distributed across 5 repositories in Transformers, GGUF and GGUF (imatrix) formats, totalling 5.1K downloads.
Specification
- Model name
- Ko-Llama-3.1-8B-Lexi-Uncensored-V2
- Source repository
- s5ya/Ko-Llama-3.1-8B-Lexi-Uncensored-V2
- Base model
- beomi/Llama-3-Open-Ko-8B
- Publisher
- s5ya
- Parameters
- 8B
- VRAM required
- ~5.6 GB at Q4_K_M, 4k context (estimated) — runs on 6 GB cards
- Hardware class
- 4B – 10B — single consumer GPU
- Licence
- llama3.1
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GGUF, GGUF (imatrix)
- First indexed
- 30 Aug 2024
- Last updated
- 8 Sept 2024
Ablation technique
How Ko-Llama-3.1-8B-Lexi-Uncensored-V2 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 is a merge of pre-trained language models created using mergekit.”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download mradermacher/Ko-Llama-3.1-8B-Lexi-Uncensored-V2-i1-GGUF --local-dir ./ko-llama-3.1-8b-lexi-uncensored-v2
llama-cli -m ./ko-llama-3.1-8b-lexi-uncensored-v2/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "s5ya/Ko-Llama-3.1-8B-Lexi-Uncensored-V2"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve s5ya/Ko-Llama-3.1-8B-Lexi-Uncensored-V2 --trust-remote-codeDownloads and variants (4)
Every published repository of Ko-Llama-3.1-8B-Lexi-Uncensored-V2, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| mradermacher/Ko-Llama-3.1-8B-Lexi-Uncensored-V2-i1-GGUF | GGUF (imatrix) | 3.4K |
| mradermacher/Ko-Llama-3.1-8B-Lexi-Uncensored-V2-GGUF | GGUF | 876 |
| QuantFactory/Ko-Llama-3.1-8B-Lexi-Uncensored-V2-GGUF | GGUF | 717 |
| s5ya/Ko-Llama-3.1-8B-Lexi-Uncensored-V2source | Transformers | 26 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- Ko-Llama-3.1-8B-Lexi-Uncensored-V2 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 beomi/Llama-3-Open-Ko-8B, not added by ablation — at 8B parameters, expect a small model’s reasoning and a high error rate on exploit detail.
- 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 llama3.1, but the terms that bind you are beomi/Llama-3-Open-Ko-8B’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 Ko-Llama-3.1-8B-Lexi-Uncensored-V2 is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with s5ya, and does not independently verify publisher claims. See responsible use.