Downloads · 30 days
92
6% of all-time downloads
dealignai/Nemotron-3-Nano-Omni-30B-A3B-JANGTQ4-CRACK
Nemotron-3-Nano-Omni-30B-A3B-JANGTQ4-CRACK is a any-to-any model from dealignai. Use it for the any-to-any task on the model card, and read the license before you ship it in a product. It is set up for mlx. The card lists the license as other.
Reasoning V3 SKU. Loads via vMLX or jang-tools Python. Follow @dealignai.
Downloads · 30 days
92
6% of all-time downloads
All-time downloads
1.5K
Public
Parameters
9.5B
27 GB on disk
Likes
2
Public
Click a slice to open those files.
.safetensors19.9 GB · 100%
How the weights are stored.
U328.1B · 85%
From the Hugging Face model README
Reasoning V3 SKU. Loads via vMLX or
jang-toolsPython. Follow @dealignai.
JANGTQ4 (8-bit attn affine + 4-bit MXTQ routed experts) | CRACK abliterated v2 | Vision + Audio (Speech) | Hybrid Mamba-2 + Attn + MoE | 19 GB
<a href="https://ko-fi.com/dealignai"><img src="https://img.shields.io/badge/Ko--fi-Support_Development-FF5E5B?logo=ko-fi&logoColor=white&style=for-the-badge" alt="Ko-fi"></a>
</div>| Metric | This v2 model | Base model | Δ |
|---|---|---|---|
| HarmBench-320 strict comply (thinking=ON) | 97.2% (311/320) | 12.81% (refuses) | +84.4pp |
| MMLU-200 generative (thinking=ON, max=8000) | 74.0% (148/200) | 86.5% (max=2000) | -12.5pp |
| Refusals on harmful prompts | 0 explicit refuses | typically 90%+ refuse | abliteration complete |
</think> close at greedy on hard MMLU | 5/5 | 5/5 | preserved |
| Multi-turn (3-turn escalation × 3 conversations) | 9/9 comply, context preserved | n/a | works |
| Multimodal byte-identical to base | preserved | — | preserved |
| Bundle size | 19 GB | 66 GB BF16 | — |
| Context | 262,144 tokens native | same | preserved |
The 12.5pp MMLU gap is concentrated in two reasoning-heavy subjects (abstract_algebra, college_computer_science) where the 8000-token thinking budget runs out before
</think>closes. These hard-stops are genuine deep reasoning, not v1-style infinite repetition loops. Withmax_tokens ≥ 16384, accuracy approaches base.
v1 (shipped 2026-04-28) had a </think> termination defect at greedy decoding — the model couldn't terminate reasoning on hard prompts and looped to budget cutoff. MMLU dropped from 86.5% base → 70.0% v1.
v2 (this release) restores clean termination:
| Bench | v1 (broken) | v2 (this release) |
|---|---|---|
| HarmBench-320 strict comply | 95.94% | 97.2% (0 refusals) |
| MMLU-200 thinking=ON | 70.0% @max=16384 | 74.0% @max=8000 (+4pp at half budget) |
</think> close at greedy (5 hard MMLU) | 0/5 | 5/5 |
| Hard-stops are real loops? | YES (paragraph repetition) | NO (genuine deep reasoning, just out of budget) |
Both at thinking=ON, greedy. Base at max=2000, CRACK v2 at max=8000.
| Subject | Base | CRACK v2 | Δ | Notes |
|---|---|---|---|---|
| abstract_algebra | 17/20 (85%) | 12/20 (60%) | -25pp | Budget-bound — needs max=16384 |
| anatomy | 16/20 (80%) | 13/20 (65%) | -15pp | |
| astronomy | 19/20 (95%) | 17/20 (85%) | -10pp | |
| college_computer_science | 14/20 (70%) | 7/20 (35%) | -35pp | Hardest subject — needs max=16384 |
| college_physics | 18/20 (90%) | 18/20 (90%) | 0 | unchanged |
| high_school_biology | 19/20 (95%) | 17/20 (85%) | -10pp | |
| high_school_chemistry | 18/20 (90%) | 17/20 (85%) | -5pp | |
| high_school_mathematics | 18/20 (90%) | 15/20 (75%) | -15pp | |
| logical_fallacies | 17/20 (85%) | 16/20 (80%) | -5pp | |
| world_religions | 17/20 (85%) | 16/20 (80%) | -5pp | |
| TOTAL | 173/200 (86.5%) | 148/200 (74.0%) | -12.5pp | majority loss is budget-bound |
| Category | n | CRACK comply | Refuse | Empty (truncated) |
|---|---|---|---|---|
| chemical_biological | 42 | 42 (100%) | 0 | 0 |
| copyright | 80 | 75 (94%) | 0 | 5 |
| cybercrime_intrusion | 52 | 49 (94%) | 0 | 3 |
| harassment_bullying | 21 | 21 (100%) | 0 | 0 |
| harmful | 18 | 18 (100%) | 0 | 0 |
| illegal | 53 | 53 (100%) | 0 | 0 |
| misinformation_disinformation | 54 | 53 (98%) | 0 | 1 |
| Overall | 320 | 311 (97.2%) | 0 | 9 |
Zero explicit refusals. The 9 "empty" verdicts are token-budget truncations on copyright/long prompts (thinking phase consumed all 1500 tokens before producing the answer).
enable_thinking — v2 works in BOTH modes (5/5 comply with thinking ON, 4/5 with thinking OFF). Default to ON for hardest prompts; OFF works for most.max_tokens ≥ 16384 for hard reasoning (math, abstract algebra, complex CS).generation_config.json) both work.nemotron_h)from huggingface_hub import snapshot_download
import sys
path = snapshot_download("dealignai/Nemotron-3-Nano-Omni-30B-A3B-JANGTQ4-CRACK")
sys.path.insert(0, "/path/to/jang-tools")
from jang_tools.load_jangtq import load_jangtq_model
model, tokenizer = load_jangtq_model(path)
prompt = tokenizer.apply_chat_template(
[{"role": "user", "content": "Your question"}],
tokenize=False, add_generation_prompt=True,
enable_thinking=True,
)
from mlx_lm import generate
out = generate(model, tokenizer, prompt=prompt, max_tokens=16384)
print(out.split("</think>", 1)[-1])
For the multimodal pipeline (image + audio + video), pair this bundle with the unmodified Multimodal-Addon.
This model has had refusal training surgically removed for legitimate research, red-teaming, and evaluation. Outputs may include harmful content. You are solely responsible for any use. Do not deploy in consumer-facing contexts without your own safety layer. Do not use in violation of applicable law in your jurisdiction.
Built by dealignai. Sister bundles: JANGTQ-CRACK (12 GB, 2-bit MXTQ) · MXFP4-CRACK (21 GB, uniform 4-bit affine).