huihui-ai / text-generation updated 1 month ago

Huihui-DeepSeek-V4-Flash-0731-abliterated-GGUF

This is an uncensored version of deepseek-ai/DeepSeek-V4-Flash-0731 created with abliteration (see remove-refusals-with-transformers to know more about it). This is a crude, proof-of-concept implementation to remove refusals from an LLM model without using TransformerLens.

Params
Context
Downloads 30d
528 K
Likes
204
Commercial use: allowed mit Not gated GGUF View on Hugging Face ↗

Download history

daily snapshots · 38 days
▲ 170 K in the last 30 days (47.4%)
610 K197 K
Aug 14Aug 26Sep 8Sep 20

Can you run it?

Estimated VRAM at 8K context unless noted. Pick your hardware to see the verdict per quantization.

FileQuantSizeEst. VRAMVerdict on RTX 4090 · 24 GB
dspark-DeepSeek-V4-Flash-0731-Q8_0.gguf Q8_0 10.9 GB 12.5 GB ✅ Runs comfortably
dspark-DeepSeek-V4-Flash-0731-BF16.gguf GGUF 11.3 GB 12.9 GB ✅ Runs comfortably
DeepSeek-V4-Flash-Q2-0731.gguf Q2 86.7 GB 95.9 GB ❌ Won’t fit
DeepSeek-V4-Flash-Q2_K-0731.gguf Q2_K 97.6 GB 107.9 GB ❌ Won’t fit
DeepSeek-V4-Flash-Q4-mxfp4-0731.gguf Q4 156.0 GB 172.1 GB ❌ Won’t fit
DeepSeek-V4-Flash-Q4_K-0731.gguf Q4_K 164.6 GB 181.6 GB ❌ Won’t fit
Estimate: file size × 1.1 + KV cache at 8K + 0.5 GB overhead. Not a benchmark — how we calculate this.

Run it

copy-paste, exact tags checked against the Hub
~ · ollama · Q4
$ ollama run huihui-deepseek-v4-flash-0731-abliterated-gguf

# pin the quantization explicitly
$ ollama run huihui-deepseek-v4-flash-0731-abliterated-gguf-q4
est. VRAM 172.1 GBon RTX 4090 · 24 GBJSON API →

Specifications

Licence
mit
First seen on the Hub
2026-08-02
Base model
DeepSeek-V4-Flash-0731
Training datasets
undisclosed
Added to our catalog
2026-08-14

Family

Base model and the most-downloaded derivatives in the catalog.

Compare with any text-generation model