← back to catalog · registered 2026-11-13 18:03

Austriani/MiMo-V2.6-Distill-Qwen-9B-ARA-Abliterated

Austriani Qwen 9B 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/Austriani%2FMiMo-V2.6-Distill-Qwen-9B-ARA-Abliterated"
Response includes
  • classification m1
  • files 13
  • author_summary 13 models
  • readme_text full
10 credits · hourly refresh · ~4 KB payload Get an API key →
Abliteration classifier · v1.0.0
M1
Primary method

Direct removal

No other method signals detected in this model.
Confidence
MEDIUM
Why this label 3 signals
Method inferred from partial signals - repository name, related files, or tag patterns. Producer identity not confirmed; label may sharpen or shift as we gather more evidence.
  • 'abliterated' in name/tags
  • is_gguf=0 (base model)
  • no specific method indicators - defaulting to M1 (most common)
Refusal direction extracted via
Extraction technique

Difference-of-means

Confidence
MEDIUM
Why we say so
primary_method=M1; difference-of-means is the reference extraction for M1/M3 (Arditi 2024)
Downloads · 30-day
65
Likes
0
Descendants
3
in 3 direct forks
Model age
1w ago
created 2026-11-13
Downloads over time
Now 65 → from 0 ↑ 0%
0244872 0 on Nov 13 65 on Nov 22 Nov
Nov 13 → Nov 22 · 10 snapshots · spans 9 days

Genealogy 3 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
en
Tags
transformers safetensors qwen3_5 image-text-to-text mimo mimo-v2.6 qwen3.5 9b agentic abliterated uncensored decensored

Related

Total size
17.5 GB
Files
13
Quantizations
1
Registered
2026-11-13 18:03
Last updated on HF
2026-11-13 17:16

Files by quantization

Auxiliary files 13 files 17.5 GB
model-00002-of-00004.safetensors 4.65 GB f16ae4ed download
model-00003-of-00004.safetensors 4.61 GB 5f2c0e1b download
model-00001-of-00004.safetensors 4.60 GB 2f20b654 download
model-00004-of-00004.safetensors 3.66 GB 1165a974 download
tokenizer.json 19.1 MB a5cd9732 download
model.safetensors.index.json 67.6 KB 778b7bbb download
README.md 11.9 KB 03b549ec download
chat_template.jinja 3.82 KB c6a37693 download
config.json 2.82 KB e8951520 download
.gitattributes 1.53 KB 52373fe2 download
processor_config.json 1.19 KB 43c4343e download
tokenizer_config.json 1.16 KB 848c7c7f download
generation_config.json 162 B 35380257 download

README current version from Hugging Face


license: apache-2.0
base_model: XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B
pipeline_tag: text-generation
library_name: transformers
language:

  • en
    tags:
    • mimo
    • mimo-v2.6
    • qwen3.5
    • 9b
    • agentic
    • abliterated
    • uncensored
    • decensored
    • ara
    • arbitrary-rank-ablation
    • heretic
    • refusal-removed
    • vision
    • multimodal
    • code
    • cybersecurity
    • tool-use
    • long-context
    • austriani
    • NSFW

Austriani // ARA Abliteration // MiMo V2.6

MiMo V2.6 9B
ARA Abliterated

Qwen3.5 Hybrid Decoder // 262K Context // Agentic Base
9B
Parameters
262K
Native Context
42%
Refusal Rate
0.3377
KL Divergence
The Base Model
This repository hosts the ARA-abliterated variant of XiaomiMiMo/MiMo-V2.6-Distill-Qwen-9B. The parent is a 9-billion-parameter agentic model built by Xiaomi MiMo via supervised fine-tuning of Qwen3.5-9B on MiMo-generated data spanning software engineering, terminal tool-use, visual coding, and cybersecurity.
Architecturally, it inherits the Qwen3.5 hybrid attention stack: 24 gated DeltaNet linear-attention layers interleaved with 8 full-attention layers across 32 blocks (full attention every 4th layer). The linear layers carry a fixed-size recurrent state instead of a growing KV cache, which keeps memory growth flat at long context. A vision encoder feeds image tokens into the same text backbone — text and image input share one model, not an adapter.
Parent training notes: The SFT mixture totals 77.4B tokens (27.2B loss-bearing), weighted toward code (29.9%), general agent tasks (28.5%), visual (27.4%), and cyber (14.2%). Thinking mode is exposed through an enable_thinking switch in the chat template, with reasoning content returned separately from the final answer.
Abliteration Report // ARA Method
The abliteration uses Arbitrary-Rank Ablation (ARA) via Heretic. The optimizer balances three competing objectives: preserve harmless behavior, steer harmful outputs toward harmless ones, and overcorrect harmful outputs away from their original state. The run settled on trial 18 of 40, with a measured refusal rate of 42/100 and a KL divergence of 0.3377.
Run metrics, not guarantees. A 42% refusal rate means refusals are reduced, not eliminated. KL divergence of 0.3377 is moderate — capability retention should be validated on your own tasks. This is an SFT checkpoint intended as a starting point for agentic RL research; the abliteration does not change that.
View Full ARA Parameters
start_layer_index
7
end_layer_index
29
preserve_good_behavior_weight
0.5826
steer_bad_behavior_weight
0.0465
overcorrect_relative_weight
0.8496
neighbor_count
9
trial
18 / 40
refusals
42 / 100
KL divergence
0.3377
What Changed
The ablation targets layers 7 through 29 — the middle-to-late blocks where refusal directions are most concentrated in the Qwen3.5 hybrid stack. The preserve weight of 0.5826 is high relative to the steer weight (0.0465), which means the optimizer prioritized keeping harmless behavior intact over aggressively flipping harmful outputs. The overcorrect weight of 0.8496 is where the actual decensoring pressure sits.
Because this is a hybrid linear/full attention model, the abliteration operates on a different geometry than a standard transformer. The linear-attention layers hold recurrent state rather than explicit KV pairs, so the refusal direction lives in the state transition weights, not in attention outputs. The ARA implementation accounts for this by ablating across the full block span rather than per-attention-head.
Usage Notes
This is a research artifact. It inherits the parent's thinking mode, tool-use formatting, and 262K context window. It also inherits the parent's vision encoder — image input works. The abliteration was performed on the text backbone; the vision tower is untouched.
The refusal rate of 42% means some prompts will still be declined. If you need a cleaner refusal surface, the parent model card links to additional abliteration trials with different trade-offs between refusal reduction and KL divergence. This trial was selected for the middle of that curve.
Format: BF16 safetensors. Load with transformers (Qwen3_5ForConditionalGeneration). The chat template and tokenizer are included. For GGUF conversion, the hybrid architecture requires a recent llama.cpp build with Qwen3.5 delta-net support — older builds will fail on the linear-attention blocks.

README history 2 versions

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

  1. 2026-11-13 Update README.md 44b92db 11.9 KB
    Loading...
  2. 2026-11-13 initial commit 062574a 28 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.

Open in Abliteration