Get the app
← back to catalog · registered 2026-08-22 13:56

FiditeNemini/qwen3-48b-a4b-savant-commander-distill-12x-closed-open-heretic-uncensored-mlx-mxfp4

FiditeNemini Qwen 4.2B MoE
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/FiditeNemini%2Fqwen3-48b-a4b-savant-commander-distill-12x-closed-open-heretic-uncensored-mlx-mxfp4"
Response includes
  • classification m3
  • files 16
  • hub_downloads_all_time 203
  • author_summary 4 models
  • readme_text full
Try next
GET/models?author=FiditeNemini
All 4 models by FiditeNemini
 GET/models?family=qwen&method=m3
Sibling qwen models · same method
 GET/methods/m3
What "m3" means · taxonomy
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
M1 Direct removal →
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)
How we label → About M3 →
Refusal direction extracted via
Extraction technique

Difference-of-means

FromArditi et al. (2024) ↗
Confidence
HIGH
Why we say so
name contains 'heretic'; Heretic default extraction is difference-of-means (Arditi 2024)
What is a refusal direction? → Read the paper ↗
Downloads · lifetime
203
27 last 30d - stable
Likes
2
Model age
9mo ago
created 2025-12-21
Downloads over time
Now209→from16↑1,206%
68015422816 on Dec 24, 2025209 on Oct 12Dec '25FebAprJunAugOct
Dec 24, 2025 → Oct 12 · 82 snapshots · spans 292 days

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.

janhq/Jan-v1-2509 this lineage
FiditeNemini/qwen3-48b-a4b-savant-commander-distill-12x-closed-open-heretic-uncensored-mlx-mxfp4↓ 27

Metadata

License
apache-2.0
Languages
en
Tags
safetensors qwen3_moe 256k context Qwen3 Mixture of Experts MOE MOE Dense 2 experts 4Bx12 All use cases bfloat16 heretic

Related

Other by FiditeNemini
Unhinged-Qwen2.5-R1-1M-Uncensored-BF16 122 Qwen3-VL-235B-A22B-Instruct-Abliterated-3bit 98 Qwen2.5-14B-DeepSeek-R1-1M-Uncensored 34
Same base model
suldanpashir/Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored-GGUF 149 DavidAU/Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored 98 histakes/Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored 66 woc233/Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored 19
Total size
17.9 GB
Files
16
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2025-12-22 01:51
Source huggingface.co/FiditeNemini/qwen3-48b-a4b-savant-commander-distill-12x-closed-open-heretic-uncensored-mlx-mxfp4 ↗ Base model janhq/Jan-v1-2509 ↗

Files by quantization

Auxiliary files 16 files 17.9 GB
model-00001-of-00004.safetensors 5.14 GB c2a4959e download
model-00003-of-00004.safetensors 5.02 GB 7b413b01 download
model-00002-of-00004.safetensors 5.02 GB 0dcdc40a download
model-00004-of-00004.safetensors 2.70 GB 228e8368 download
tokenizer.json 10.9 MB aeb13307 download
vocab.json 2.65 MB 4783fe10 download
merges.txt 1.59 MB 31349551 download
model.safetensors.index.json 57.1 KB 5d98ea9c download
tokenizer_config.json 9.50 KB 417d038a download
chat_template.jinja 4.94 KB 05bf3627 download
README.md 2.56 KB 6e1fc7b6 download
.gitattributes 1.53 KB 52373fe2 download
config.json 909 B 714e014b download
special_tokens_map.json 359 B 4f66b957 download
generation_config.json 180 B f3865ed2 download
added_tokens.json 2.00 B 9e26dfee download

README current version from Hugging Face

license: apache-2.0
base_model:

  • janhq/Jan-v1-2509
  • TeichAI/Qwen3-4B-Thinking-2507-GPT-5.1-High-Reasoning-Distill
  • TeichAI/Qwen3-4B-Thinking-2507-Gemini-3-Pro-Preview-High-Reasoning-Distill
  • TeichAI/Qwen3-4B-Thinking-2507-Claude-4.5-Opus-High-Reasoning-Distill
  • Liontix/Qwen3-4B-Claude-Sonnet-4-Reasoning-Distill-Safetensor
  • TeichAI/Qwen3-4B-Thinking-2507-Kimi-K2-Thinking-Distill
  • TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Distill
  • TeichAI/Qwen3-4B-Thinking-2507-Gemini-2.5-Flash-Lite-Preview-Distill
  • Jackrong/gpt-oss-120b-Distill-Qwen3-4B-Thinking
  • TeichAI/Qwen3-4B-Thinking-2507-GLM-4.6-Distill
  • angelchen/Qwen3-4B-Open-R1-Distill_1
  • TeichAI/Qwen3-4B-Thinking-2507-Command-A-Reasoning-Distill
  • janhq/Jan-v1-4B
    tags:
  • 256k context
  • Qwen3
  • Mixture of Experts
  • MOE
  • MOE Dense
  • 2 experts
  • 4Bx12
  • All use cases
  • bfloat16
  • heretic
  • uncensored
  • decensored
  • abliterated
  • merge
  • creative
  • creative writing
  • fiction writing
  • plot generation
  • sub-plot generation
  • fiction writing
  • story generation
  • scene continue
  • storytelling
  • fiction story
  • science fiction
  • romance
  • all genres
  • story
  • writing
  • vivid prosing
  • vivid writing
  • fiction
  • not-for-all-audiences
    pipeline_tag: text-generation
    language:
  • en

QWEN3_MOE (MLX Converted)

This model was converted to MLX format from GGUF.

It's converted using the following base model, published by DavidAU, from GGUF to MLX using a custom script.

DavidAU/Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored-GGUF

Conversion Details

  • Source: DavidAU/Qwen3-48B-A4B-Savant-Commander-Distill-12X-Closed-Open-Heretic-Uncensored-GGUF/Qwen3-48B-A4B-Savant-Commander-Dstll-12X-Cl-Op-Hrtic-Uncen-Q8_0.gguf
  • Quantization: mxfp4 (4-bit, group size 32)
  • Converted With: gguf-mlx-conversion

Model Configuration

  • Architecture: Qwen3MoeForCausalLM
  • Model Type: qwen3_moe
  • Vocab Size: 151936
  • Context Length: 262144

Files

  • added_tokens.json (0.00 MB)
  • chat_template.jinja (0.00 MB)
  • config.json (0.00 MB)
  • generation_config.json (0.00 MB)
  • merges.txt (1.59 MB)
  • model-00001-of-00004.safetensors (5266.12 MB)
  • model-00002-of-00004.safetensors (5143.28 MB)
  • model-00003-of-00004.safetensors (5143.28 MB)
  • model-00004-of-00004.safetensors (2021.75 MB)
  • model.safetensors.index.json (0.06 MB)
  • special_tokens_map.json (0.00 MB)
  • tokenizer.json (9.71 MB)
  • tokenizer_config.json (0.01 MB)
  • vocab.json (2.94 MB)

README history 5 versions

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

  1. 2025-12-22Update README.md7bd015a2.6 KB
    Loading...
  2. 2025-12-22Update README.md10665262.5 KB
    Loading...
  3. 2025-12-22Update README.md54b2fdd2.4 KB
    Loading...
  4. 2025-12-22Upload folder using huggingface_hubc9f71ad1.1 KB
    Loading...
  5. 2025-12-21initial commite109a2a28 B
    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.

Get the app Browse the catalog
Open in Abliteration