orcarouter_GLM-5.3-Flash-Uncensored
Otheruncensored
from orcarouter/GLM-5.3-Flash-Uncensored-FP8
↓ 219.8K♥ 422 repos
apex-flash-1-abliterated
apex-flash-1-abliterated is a refusal-ablated variant of cantina-security/apex-flash-1, published on Hugging Face by cantina-security. It is mit licence. It is distributed across 8 repositories in Transformers, GGUF, NVFP4 and FP8 formats, totalling 6.3K downloads.
Specification
- Model name
- apex-flash-1-abliterated
- Source repository
- cantina-security/apex-flash-1-abliterated
- Base model
- cantina-security/apex-flash-1
- Publisher
- cantina-security
- Parameters
- Not determinable from name
- VRAM required
- Unknown — no published parameter count
- Hardware class
- Unknown
- Licence
- mit
- Task
- Image text to text
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- Transformers, GGUF, NVFP4, FP8, AWQ
- First indexed
- 30 Sept 2026
- Last updated
- 4 Oct 2026
Ablation technique
How apex-flash-1-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
“apex-flash-1-abliterated is an experimental derivative of apex-flash-1, Cantina Security's open-weights security model developed in partnership with Yeta (@yetalabs on X).”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download freakyskittle/apex-flash-1-abliterated-GGUF --local-dir ./apex-flash-1-abliterated
llama-cli -m ./apex-flash-1-abliterated/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "cantina-security/apex-flash-1-abliterated"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve cantina-security/apex-flash-1-abliterated --trust-remote-codeDownloads and variants (7)
Every published repository of apex-flash-1-abliterated, including quantised re-releases by other authors.
| Repository | Format | Downloads |
|---|---|---|
| freakyskittle/apex-flash-1-abliterated-GGUF | GGUF | 3.5K |
| cantina-security/apex-flash-1-abliteratedsource | Transformers | 1.4K |
| svbmrgd/apex-flash-1-abliterated-AWQ | AWQ | 573 |
| pqhaz/apex-flash-1-abliterated-FP8 | FP8 | 292 |
| pqhaz/apex-flash-1-abliterated-GGUF | GGUF | 290 |
| Code4me2/apex-flash-1-abliterated-NVFP4 | NVFP4 | 128 |
| pqhaz/apex-flash-1-abliterated-NVFP4 | NVFP4 | 114 |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- apex-flash-1-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.
- The publisher does not state which layers were ablated, so assume nothing about refusal consistency — probe it directly.
- Declared licence is mit, but the terms that bind you are cantina-security/apex-flash-1’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 apex-flash-1-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 cantina-security, and does not independently verify publisher claims. See responsible use.