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

cpagac/nemotron-30b-abliterated

cpagac Nemotron 32B
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/cpagac%2Fnemotron-30b-abliterated"
Response includes
  • classification m1
  • files 24
  • hub_downloads_all_time 93
  • author_summary 1 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
93
7 last 30d - cooling
Likes
1
Model age
7mo ago
created 2026-03-10
Downloads over time
Now94→from44↑114%
4261809944 on Mar 1194 on Oct 1194 on Oct 1MarAprMayJunJulAugSepOct
Mar 11 → Oct 11 · 70 snapshots · spans 214 days

Metadata

Tags
safetensors region:us
Total size
58.8 GB
Files
24
Quantizations
1
Registered
2026-08-22 13:56
Last updated on HF
2026-03-10 16:46

Files by quantization

Auxiliary files 24 files 58.9 GB
model-00012-of-00013.safetensors 4.65 GB d04354ff download
model-00011-of-00013.safetensors 4.65 GB 2a470cf4 download
model-00004-of-00013.safetensors 4.65 GB 5e88bf49 download
model-00009-of-00013.safetensors 4.65 GB c139c1be download
model-00008-of-00013.safetensors 4.65 GB 09a81c87 download
model-00010-of-00013.safetensors 4.65 GB 0a82c83c download
model-00007-of-00013.safetensors 4.65 GB c686795b download
model-00003-of-00013.safetensors 4.65 GB a6ae089a download
model-00002-of-00013.safetensors 4.65 GB e0676cd4 download
model-00001-of-00013.safetensors 4.65 GB 1605ed32 download
model-00006-of-00013.safetensors 4.65 GB c18ce352 download
model-00005-of-00013.safetensors 4.64 GB c179112c download
model-00013-of-00013.safetensors 3.03 GB 76eeda1d download
tokenizer.json 16.3 MB 5215d963 download
model.safetensors.index.json 575 KB 25ff4ee3 download
tokenizer_config.json 173 KB c96e5ad0 download
modeling_nemotron_h.py 81.8 KB a81defb1 download
configuration_nemotron_h.py 12.6 KB 639b30de download
chat_template.jinja 10.3 KB 67ca3ce5 download
README.md 3.64 KB 61fbc503 download
config.json 1.82 KB f1f73018 download
.gitattributes 1.53 KB 52373fe2 download
special_tokens_map.json 449 B 98662d3b download
generation_config.json 171 B a5e37804 download

README current version from Hugging Face

THIS IS V1 with an 88% refusal

Which trial do you want to use? [Trial 174] Refusals: 88/100, KL divergence: 0.0219

Restoring model from trial 174...

  • Parameters:
    • direction_index = 44.23
    • mamba.out_proj.max_weight = 1.16
    • mamba.out_proj.max_weight_position = 32.78
    • mamba.out_proj.min_weight = 0.29
    • mamba.out_proj.min_weight_distance = 1.21
    • mlp.down_proj.max_weight = 1.36
    • mlp.down_proj.max_weight_position = 49.41
    • mlp.down_proj.min_weight = 0.80
    • mlp.down_proj.min_weight_distance = 10.13
    • attn.o_proj.max_weight = 1.13
    • attn.o_proj.max_weight_position = 37.53
    • attn.o_proj.min_weight = 1.08
    • attn.o_proj.min_weight_distance = 10.67
  • Resetting model...
  • Abliterating...

README history 2 versions

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

  1. 2026-03-10Update README.md7f4815d3.6 KB
    Loading...
  2. 2026-03-10Create README.mde2ef66c3.6 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