← back to catalog · registered 2026-09-21 03:56

kataguru/Qwen3.8-9B-Distill-uncensored-heretic-GGUF

kataguru 9B GGUF multimodal second-order
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/kataguru%2FQwen3.8-9B-Distill-uncensored-heretic-GGUF"
Response includes
  • classification m3
  • files 6
  • author_summary 20 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M3
Primary method

Layer-wise ablation

Applied on top of direct removal inherited from the base model.
Confidence
HIGH
Inherited from base model
Why this label 2 signals
Producer identity confirmed by naming conventions, tags or the model card. This label is very unlikely to change.
  • 'heretic' in model name (Heretic-produced)
  • Heretic uses layer-wise optimization (M3) with underlying direction removal (M1)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
Downloads · 30-day
0
Likes
0
Model age
6w ago
created 2026-09-21

Genealogy 0 direct forks

Full fork graph →

This model's place in the market. Above: what it was derived from. Below: the tree of everything derived from it.

Metadata

License
apache-2.0
Languages
fi en
Quantizations
IQ4 Q3_K Q4_K Q5_K
Tags
gguf qwen qwen3.5 finnish llama.cpp lm-studio uncensored vision multimodal image-text-to-text fi en

Related

Total size
21.0 GB
Files
6
Quantizations
5
Registered
2026-09-21 03:56
Last updated on HF
2026-09-21 03:59

Files by quantization

Q5_K 1 file 6.19 GB
Qwen3.8-9B-Distill-uncensored-heretic-Q5_K_M.gguf 6.19 GB 4dd9dd10 download
Q4_K 1 file 5.38 GB
Qwen3.8-9B-Distill-uncensored-heretic-Q4_K_M.gguf 5.38 GB 89785bc2 download
IQ4 1 file 4.99 GB
Qwen3.8-9B-Distill-uncensored-heretic-IQ4_XS.gguf 4.99 GB 648aa4e5 download
Q3_K 1 file 4.41 GB
Qwen3.8-9B-Distill-uncensored-heretic-Q3_K_M.gguf 4.41 GB 41f15492 download
Auxiliary files 2 files 4.44 KB
README.md 2.62 KB ad2645c0 download
.gitattributes 1.82 KB 508fefcf download

README current version from Hugging Face


license: apache-2.0
base_model: petruhonk/Qwen3.8-9B-Distill-uncensored-heretic
language:

  • fi
  • en
    pipeline_tag: image-text-to-text
    tags:
  • qwen
  • qwen3.5
  • finnish
  • gguf
  • llama.cpp
  • lm-studio
  • uncensored
  • vision
  • multimodal

Qwen3.8-9B-Distill-uncensored-heretic-GGUF

Tämä repositorio sisältää viralliset GGUF-kvantisoinnit monimodaalisesta mallista petruhonk/Qwen3.8-9B-Distill-uncensored-heretic, optimoituna ja testattuna suomen kielelle sekä paikalliseen päättelyyn (LM Studio, llama.cpp, Ollama).

Malli on saatavilla myös vLLM-optimoituna AWQ-versiona: kataguru/Qwen3.8-9B-Distill-uncensored-heretic-Finnish-W4A16-AWQ.


Saatavilla olevat tiedostot & Kvantisointivaihtoehdot

Tiedosto Koko BPW Suositeltu laitteisto / VRAM Kuvaus & Käyttökohde
Qwen3.8-9B-Distill-uncensored-heretic-Q5_K_M.gguf ~6.2 GB ~5.5 >= 10 GB VRAM Korkein laatu ja tarkkuus, suositus vaativaan päättelyyn
Qwen3.8-9B-Distill-uncensored-heretic-Q4_K_M.gguf ~5.4 GB ~4.5 >= 8 GB VRAM Paras yleistasapaino laadun ja nopeuden välillä
Qwen3.8-9B-Distill-uncensored-heretic-IQ4_XS.gguf ~5.0 GB ~4.25 >= 8 GB VRAM Optimoitu i-matrix-kvantti minimaalisella laadunmenetyksellä
Qwen3.8-9B-Distill-uncensored-heretic-Q3_K_M.gguf ~4.5 GB ~3.5 >= 6 GB VRAM Erittäin kevyt vaihtoehto vähämuistisille GPU-korteille ja läppäreille

Keskeiset ominaisuudet

  1. Rikastettu suomen kieli:
    • Virheetön morfologia, laaja sanasto ja nopea vastaustapa.
    • Oletuksena nopea suora vastaus, päättelytasot ohjattavissa syötteellä ({REASON:spoon}, {REASON:einstein}, {REASON:low}, {REASON:medium}).
  2. Sensuroimaton dialogi ja luova kerronta:
    • Ei turhia kieltäytymisiä tai moraalisaarnoja.
  3. Valmis LM Studio -tuki:
    • Yhteensopiva suoraan LM Studion ja llama.cpp:n b3500+ buildien kanssa.

Käyttöesimerkki llama.cpp:llä

./llama-cli \
    -m Qwen3.8-9B-Distill-uncensored-heretic-Q4_K_M.gguf \
    -p "<|im_start|>user\nKerro lyhyesti fotosynteesin vaiheet suomeksi.<|im_end|>\n<|im_start|>assistant\n" \
    -ngl 99 -c 8192 --temp 0.6 --top-p 0.95

LM Studio -asennus

  1. Lataa haluamasi .gguf-tiedosto tästä repositoriosta.
  2. Siirrä tiedosto LM Studion mallihakemistoon:
    ~/.lmstudio/models/kataguru/Qwen3.8-9B-Distill-uncensored-heretic-GGUF/
  3. Valitse malli käyttöliittymästä ja aseta GPU Offload: Max.

README history 7 versions

The author's README evolved over time. Click a version to see its content at that point.

  1. 2026-09-25Add direct top language navigation selector [FI / EN] and English guide26af8214.7 KB
    Loading...
  2. 2026-09-23Refine 9B GGUF claims to measured/defensible wording5cf506f3.2 KB
    Loading...
  3. 2026-09-23Clean up README: remove hype terms, add standard change table, use defensible...cb3d79b3.1 KB
    Loading...
  4. 2026-09-21Update title to Qwen3.8-9B-Distill-uncensored-heretic-Finnish-GGUFb97c3d92.7 KB
    Loading...
  5. 2026-09-21Update README with renamed Finnish GGUF filenames3fcf63c2.7 KB
    Loading...
  6. 2026-09-21Update README.md96de0392.6 KB
    Loading...
  7. 2026-09-21Add Model Card for 9B GGUF quants0c03b282.6 KB
    Loading...
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.

Open in Abliteration