Get the app
← back to catalog · registered 2026-10-20 16:00

edp1096/Huihui-DeepSeek-V4-Flash-Vision-Exp-abliterated-exl3

edp1096 Deepseek multimodal
Your rig guess connected
? Why do I need an app?
Reading your rig…

This is a rough estimate. Install the free app - we'll show exact numbers.

Reading real hardware from your app right now. Numbers below are exact.

Below is the per-quantization compatibility for this model. 

curl -H "Authorization: Bearer $ABL_KEY" \
     "https://abliteration.org/api/v1/models/edp1096%2FHuihui-DeepSeek-V4-Flash-Vision-Exp-abliterated-exl3"
Response includes
  • classification unknown
  • files 2
  • author_summary 13 models
  • readme_text full
Try next
GET/models?author=edp1096
All 13 models by edp1096
 GET/models?family=deepseek&method=unknown
Sibling deepseek models · same method
 GET/methods/unknown
What "unknown" means · taxonomy
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
?
Primary method

Unclassified

No clear signals of an abliteration technique in this model.
Confidence
UNKNOWN
Why this label 1 signal
No classification signals present. This may not be an abliterated model at all - it could be a repackaging, a merge with unrelated goals, or unrelated content that mentions the term.
  • no classification signals present (no abliterated, uncensored, or known producer/method markers)
How we label → About ? →
Refusal direction extraction

No specific extraction method could be identified for this model. The producer either did not document it or used a proprietary pipeline.

What is a refusal direction? →
Downloads · 30-day
0
Likes
0
Model age
today
created 2026-10-20

Metadata

License
mit
Tags
exl3 huihui abliterated vision image-text-to-text license:mit region:us

Related

Other by edp1096
Huihui-RadixArk-Qwen3.8-27B-abliterated-NVFP4 4K Huihui-RadixArk-Qwen3.8-Flash-Next-abliterated-NVFP4 2K Huihui-gemma-4-E2B-it-abliterated-litert-lm 1K Huihui-Qwen3.8-Flash-Next-abliterated-NVFP4-QAD 778 Huihui-GLM-5.3-Flash-abliterated-NVFP4 371
Total size
0 B
Files
2
Quantizations
1
Registered
2026-10-20 16:00
Last updated on HF
2026-10-20 15:48
Source huggingface.co/edp1096/Huihui-DeepSeek-V4-Flash-Vision-Exp-abliterated-exl3 ↗

Files by quantization

Auxiliary files 2 files 3.24 KB
README.md 1.76 KB 895576b9 download
.gitattributes 1.48 KB a6344aac download

README current version from Hugging Face

license: mit
pipeline_tag: image-text-to-text
tags:

  • exl3
  • huihui
  • abliterated
  • vision
Branch Weights Korean TG Code TG Actual post-transplant input lengths tested
2.04bpw 75.97 GiB 33.7 40.0 129,996
2.27bpw 83.47 GiB 34.7 40.3 129,996, 1,046,999
2.52bpw 94.08 GiB 32.4 37.1 129,996

TG is tok/s on short synthetic samples with MTP 3. Every branch is configured for a 1M window; consult the last column for actual post-transplant context tests.

Tested engine: native ExLlamaV3 1.5.4 with the mixed-MLP guard included in each branch, plus TabbyAPI. Read the branch instructions before loading with the pinned stock engine.

Each checkpoint preserves its base EXL3 routed experts, vision, and MTP, replacing Huihui's 18 attention-output B and 18 shared-expert down projections. Those donor tensors are Q8_0 decoded to F16. Branch names denote the base quantization setting; they do not describe a new effective average bit rate.

Select a branch to download weights and read its detailed validation and provenance. main contains this index only.

Source

Catalog is the map. Apps are the tools.

Run models on your own machine, not in the cloud.

Every model page has an "Open in Abliteration" button that hands the model directly to the first-party desktop client, at the quantization your rig can actually run. No API keys, no subscription, no prompt leakage.

Get the app Browse the catalog
Open in Abliteration