audreyt/DeepSeek-V4-Flash-Vision-Exp-Abliterated-GGUF overview
DeepSeek V4 Flash Vision Exp Abliterated ds4 IQ2 Official recipe Vision Exp language GGUF IQ2 XXS routed experts, Q2 K down, Q8 0 attention/shared/output with …
Runs locally from ~80.76 GB disk (32 GB+ VRAM class GPUs with llama.cpp / guIDE).
Repository Files & Downloads
| File | Type | Quantization | Size | Link |
|---|---|---|---|---|
| DeepSeek-V4-Flash-Vision-Exp-Abliterated-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8.gguf | GGUF | IQ2XXS | 80.76 GB | Download |
Model Details
Model README
---
license: mit
base_model: deepseek-ai/DeepSeek-V4-Flash-Vision-Exp
tags:
- deepseek
- deepseek-v4-flash
- vision-exp
- gguf
- ds4
- abliterated
---
DeepSeek V4 Flash Vision-Exp Abliterated (ds4 IQ2)
Official-recipe Vision-Exp language GGUF (IQ2_XXS routed experts, Q2_K down, Q8_0 attention/shared/output) with the published rank-1 refusal-direction edit baked into 33 blk.{10..42}.attn_output_b.weight tensors.
This is not Headroom128 0731 Flash. Do not attach the 0731 DSpark support GGUF. Pair with the unmodified 316-tensor encoder.
Files
| file | bytes | sha256 |
|---|---:|---|
| DeepSeek-V4-Flash-Vision-Exp-Abliterated-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8.gguf | 86720111776 | 66e47437ce7201546fa870a23ccdc094282660f1e52f3403ed5fb59604b2d894 |
| encoder (not in this repo) | 932857760 | see antirez/deepseek-v4-gguf DeepSeek-V4-Flash-Vision-Encoder.gguf |
Language KV: 1328 tensors / 66 KV, architecture=deepseek4, checkpoint_variant=vision-exp, sidecar_required=true, source.revision=e46e16bf6035c6f317eb2ac7458eb0362926d402.
How it was made
- APFS-clone the official antirez Vision-Exp IQ2 language GGUF.
- HTTP-Range the 33 already-quantized
Q8_0attn_output_bpayloads fromapetersson/DeepSeek-V4-Flash-Vision-Exp-Abliteratedrevisionf41380effb3b78f9e08cec2110296a149fb2e063(Basic128-Routed-IQ2_M),Content-Rangetotal102826285248. pwriteeach payload at the parsed localdata_pos+rel_offset(not a 35651584-byte stride).- SHA-256 every destination payload against the pinned Basic128 source. Control
blk.0.attn_output_b.weightunchanged (384e5056…).
Abliteration recipe (publisher): rank-1 residual on HF layers.N.attn.wo_b for N=10..42, λ=3.5, preserve output-row L2; MTP/shared-down untouched. Direction transferred from drowzeys 0731 DSpark ablit 85ae5bba.
Run (ds4 origin/main / audreyt d0c2b43)
./ds4 -m DeepSeek-V4-Flash-Vision-Exp-Abliterated-IQ2XXS-w2Q2K-AProjQ8-SExpQ8-OutQ8.gguf \
--vision DeepSeek-V4-Flash-Vision-Encoder.gguf \
--nothink --temp 0
# then: /read photo.jpg
M5 Max 128 GiB smoke (2026-09-01): image 512x507, 169 image tokens, Earth/Africa globe caption, 285.75 prefill / 44.94 gen t/s.
License
MIT, inheriting DeepSeek V4 Flash / Vision-Exp. Derivative of deepseek-ai/DeepSeek-V4-Flash-Vision-Exp and the official antirez IQ2 quant. Abliterated wo_b payloads from apetersson Basic128 (same MIT inherit).
Run audreyt/DeepSeek-V4-Flash-Vision-Exp-Abliterated-GGUF with guIDE
Download guIDE — the AI-native code editor with local LLM inference and 69 built-in tools.
Source: Hugging Face · Compare models