DeepSeek-V4-Flash-0731-abliterated
DeepSeek
from deepseek-ai/DeepSeek-V4-Flash-0731
↓ 645.8K♥ 19023 repos
DeepSeek-V4-Flash-Strix-Halo
DeepSeek-V4-Flash-Strix-Halo is a refusal-ablated variant of deepseek-ai/DeepSeek-V4-Flash-0731, published on Hugging Face by otheru. It is other licence. It is distributed across 2 repositories in GGUF format, totalling 24.3K downloads.
Specification
- Model name
- DeepSeek-V4-Flash-Strix-Halo
- Source repository
- otheru/DeepSeek-V4-Flash-Strix-Halo-GGUF
- Base model
- deepseek-ai/DeepSeek-V4-Flash-0731
- Publisher
- otheru
- Parameters
- Not determinable from name
- Hardware class
- Unknown
- Licence
- other
- Task
- Text generation
- Context window
- Not documented
- Ablation scope
- Not stated by the publisher
- Formats available
- GGUF
- First indexed
- 26 Jul 2026
- Last updated
- 3 Sept 2026
Ablation technique
How DeepSeek-V4-Flash-Strix-Halo 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
“Measured 2026-08-22 on one AMD Ryzen AI Max+ 395 with Radeon 8060S (gfx1151) and 128 GB unified memory, against the artifacts published here (target SHA-256 a936e0a5…, drafter 1a01c80e…, both re-verified against the local copies before the run). Runtime was Ember release 2026.8.22 in ember-rocm:7.14, with speculative...”
Running it
Commands are templates — confirm the exact repository and quant file before use.
llama.cpp / GGUF
hf download otheru/DeepSeek-V4-Flash-Strix-Halo-GGUF --local-dir ./deepseek-v4-flash-strix-halo
llama-cli -m ./deepseek-v4-flash-strix-halo/<file>.gguf -p "..."Transformers
from transformers import AutoModelForCausalLM, AutoTokenizer
repo = "otheru/DeepSeek-V4-Flash-Strix-Halo-GGUF"
tok = AutoTokenizer.from_pretrained(repo)
model = AutoModelForCausalLM.from_pretrained(
repo, torch_dtype="auto", device_map="auto"
)vLLM
vllm serve otheru/DeepSeek-V4-Flash-Strix-Halo-GGUF --trust-remote-codeDownloads and variants (1)
The only published repository of DeepSeek-V4-Flash-Strix-Halo.
| Repository | Format | Downloads |
|---|---|---|
| otheru/DeepSeek-V4-Flash-Strix-Halo-GGUFsource | GGUF | 24.3K |
Security-team assessment
Derived from this model’s metadata and the publisher’s claims — not from benchmarks run by this site.
- DeepSeek-V4-Flash-Strix-Halo 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 other, but the terms that bind you are deepseek-ai/DeepSeek-V4-Flash-0731’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-Strix-Halo is collected from the public Hugging Face API and the publisher’s model card. This site does not host weights, is not affiliated with otheru, and does not independently verify publisher claims. See responsible use.