DeepSeek-V4-Flash-0731-Abliterated-DS4
DeepSeek
from apetersson/DeepSeek-V4-Flash-0731-Abliterated-FP8
↓ 52.3K♥ 33
DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128
DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 is a refusal-ablated variant of apetersson/DeepSeek-V4-Flash-0731-Abliterated-FP8, published on Hugging Face by apetersson. It is mit licence. It is distributed across 2 repositories in GGUF format, totalling 16.8K downloads.
Specification
- Model name
- DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128
- Source repository
- apetersson/DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128
- Publisher
- apetersson
- Parameters
- Not determinable from name
- Hardware class
- Unknown
- Licence
- mit
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF
- First indexed
- 1 Aug 2026
- Last updated
- 4 Aug 2026
Ablation technique
How DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 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
“This is a quality-first GGUF package built with the DS4 Quality128 quantization policy. It is designed to keep DeepSeek V4 Flash resident on a 128 GB M1 Ultra while preserving the most sensitive routed experts in their exact native MXFP4 representation. DS4 in the package name identifies the quantization profile. Use...”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download apetersson/DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 --local-dir ./deepseek-v4-flash-0731-abliterated-ds4-quality128
llama-cli -m ./deepseek-v4-flash-0731-abliterated-ds4-quality128/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "apetersson/DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve apetersson/DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 --trust-remote-codeDownloads and variants (1)
The only published repository of DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128.
| Repository | Format | Downloads |
|---|---|---|
| apetersson/DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128source | GGUF | 16.8K |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 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 apetersson/DeepSeek-V4-Flash-0731-Abliterated-FP8’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 DeepSeek-V4-Flash-0731-Abliterated-DS4-Quality128 is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with apetersson, and does not independently verify publisher claims. See responsible use.