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

noneUsername/Qwen3-32B-abliterated-llm-compressor-AWQ

noneUsername Qwen 31B
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/noneUsername%2FQwen3-32B-abliterated-llm-compressor-AWQ"
Response includes
  • classification m1
  • files 17
  • hub_downloads_all_time 144
  • author_summary 17 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 · lifetime
144
9 last 30d - cooling
Likes
0
Model age
15mo ago
created 2025-06-26
Downloads over time
Now147→from6↑2,350%
0541071616 on Jul 9, 2025147 on Oct 11Jul '25Sep '25Nov '25JanMarMayJulSep
Jul 9, 2025 → Oct 11 · 105 snapshots · spans 459 days

Metadata

Tags
safetensors qwen3 compressed-tensors region:us

Related

Total size
18.0 GB
Files
17
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2025-06-26 21:09

Files by quantization

Auxiliary files 17 files 18.0 GB
model-00002-of-00004.safetensors 4.66 GB e03998fb download
model-00001-of-00004.safetensors 4.63 GB 08b06ce4 download
model-00003-of-00004.safetensors 4.61 GB a1ade85a download
model-00004-of-00004.safetensors 4.11 GB a5be5197 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 177 KB f9110284 download
tokenizer_config.json 5.28 KB ddaf6980 download
chat_template.jinja 4.02 KB 699ff8df download
README.md 3.14 KB ca44ad97 download
.gitattributes 1.53 KB 52373fe2 download
config.json 1.45 KB 70f63a03 download
added_tokens.json 707 B b54f9135 download
special_tokens_map.json 613 B ac23c0aa download
generation_config.json 160 B bf86b0a4 download
recipe.yaml 126 B 3ffbab87 download

README current version from Hugging Face

vllm (pretrained=/root/autodl-tmp/Qwen3-32B-abliterated,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,tensor_parallel_size=4,gpu_memory_utilization=0.8), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.900 ± 0.0190
strict-match 5 exact_match ↑ 0.896 ± 0.0193

vllm (pretrained=/root/autodl-tmp/Qwen3-32B-abliterated,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,tensor_parallel_size=4,gpu_memory_utilization=0.8), gen_kwargs: (None), limit: 500.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.852 ± 0.0159
strict-match 5 exact_match ↑ 0.840 ± 0.0164
Groups Version Filter n-shot Metric Value Stderr
mmlu 2 none acc ↑ 0.7988 ± 0.0131
- humanities 2 none acc ↑ 0.7897 ± 0.0269
- other 2 none acc ↑ 0.7590 ± 0.0298
- social sciences 2 none acc ↑ 0.8722 ± 0.0252
- stem 2 none acc ↑ 0.7860 ± 0.0230

vllm (pretrained=/root/autodl-tmp/Qwen3-32B-abliterated-128-4096,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.8), gen_kwargs: (None), limit: 250.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.848 ± 0.0228
strict-match 5 exact_match ↑ 0.844 ± 0.0230

vllm (pretrained=/root/autodl-tmp/Qwen3-32B-abliterated-128-4096,add_bos_token=true,max_model_len=3096,dtype=bfloat16,trust_remote_code=true,gpu_memory_utilization=0.8), gen_kwargs: (None), limit: 500.0, num_fewshot: 5, batch_size: auto

Tasks Version Filter n-shot Metric Value Stderr
gsm8k 3 flexible-extract 5 exact_match ↑ 0.850 ± 0.0160
strict-match 5 exact_match ↑ 0.834 ± 0.0167
Groups Version Filter n-shot Metric Value Stderr
mmlu 2 none acc ↑ 0.7860 ± 0.0133
- humanities 2 none acc ↑ 0.7795 ± 0.0272
- other 2 none acc ↑ 0.7538 ± 0.0295
- social sciences 2 none acc ↑ 0.8389 ± 0.0271
- stem 2 none acc ↑ 0.7789 ± 0.0232

README history 1 version

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

  1. 2025-06-26Create README.mdcf28ac23.1 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