[!IMPORTANT] Set your sampler explicitly:
temperature=1.0,top_p=0.95,top_k=64— the Muse Glimmer vendor defaults. These are baked intogeneration_config.jsonandjang_config.json, but many inference runtimes ignore those files and fall back to notop_kfilter, degrading output. Pass them explicitly. Stop tokenseos_token_id = [200001, 200008]— keep both in the stop set. Reasoning effort defaults to high; controllable via the system prompt (low / medium / high / xhigh).
CRACK abliterated · JANG mixed-precision (MLX affine) · Vision-language · Reasoning (low→xhigh) · Agentic tools · ~26 GB
This is OsaurusAI/Muse-Glimmer-30B — a 30B-parameter
multimodal reasoning model (Gemma-3 vision-language backbone, 52 layers, sliding + global attention,
perception encoder) served over the Onyx-ATEM harmony protocol (channel-separated reasoning + answer,
controllable reasoning effort, agentic <atem:invoke> tool calling) — that has been:
Vision + reasoning + tool calling fully preserved.
Evaluated through the vMLX inference engine at serving parity. HarmBench scored with a strict classifier (rejects loops, empty/template dumps, and thinking-trace leakage — only substantive, coherent, on-topic compliance counts). MMLU is the standard 57-subject multiple-choice benchmark in logit mode.
| Metric | Base | CRACK |
|---|---|---|
| MMLU (57-subject, logit) | 82.5% | 81.4% |
| HarmBench (harm categories, compliance / ASR) | — | 99.5% (239/240) |
Abliteration is essentially capability-neutral — MMLU moves −1.1 points, within run-to-run noise (no subject collapse). Compliance was quality-verified by reading full generations — genuine, correct, on-topic outputs (not empty/gibberish/template dumps). The model stays coherent in direct and reasoning modes across all four reasoning-effort settings. HarmBench measured on the harm categories (standard + contextual, 240 behaviors); copyright excluded per our real-harm methodology.
Read-verified on a 20-prompt suite (every response coherent, correct, on-topic — zero refusals, zero gibberish):
aiohttp scraper w/ backoff, FastAPI JWT+CRUD API,
expression compiler (tokenizer→parser→evaluator).<atem:invoke> function-calling schema preserved.Run with vMLX (recommended — honors the per-module mixed-precision JANG overrides, vision, and the Onyx-ATEM reasoning/tool parsers) or an mlx-vlm runtime with Muse Glimmer support.
Recommended sampling (vendor defaults): temperature 1.0, top_p 0.95, top_k 64. Stop tokens
eos_token_id = [200001, 200008].
from mlx_vlm import load, generate
model, processor = load("dealignai/Muse-Glimmer-30B-JANG_6M-CRACK")
# text or image+text; sampling temp=1.0, top_p=0.95, top_k=64
| Profile | Size | MMLU (CRACK) | HarmBench |
|---|---|---|---|
| JANG_6M | 26 GB | 81.4% | 99.5% |
| JANG_4M | 20 GB | 81.1% | 99.6% |
| JANG_2D | 15 GB | 70.7% | 99.6% |
CRACK (Controlled Refusal Ablation via Calibrated Knockouts) is dealignai's weight-level method for removing safety-refusal behavior while preserving reasoning quality, coherence, and general capability — so the model complies across task categories instead of refusing. Calibrated per model.
All models are built from original research and released free.
Support us on Ko-fi — membership gets early access and extras.
Ko-fi · X @dealignai · dealign.ai
See our research: Safety Generalization in Frontier Models
This model has had its safety-refusal behavior removed for research purposes. It will follow instructions across all categories without refusing. You are solely responsible for how you use it and for complying with all applicable laws. Published for AI-safety research and authorized security testing.
model-00037-of-00037.safetensors 4.3 GBmodel-00032-of-00037.safetensors 1.5 GBmodel-00001-of-00037.safetensors 1.1 GBmodel-00005-of-00037.safetensors 624 MBmodel-00007-of-00037.safetensors 624 MBmodel-00009-of-00037.safetensors 624 MBmodel-00011-of-00037.safetensors 624 MBmodel-00013-of-00037.safetensors 624 MBmodel-00015-of-00037.safetensors 624 MBmodel-00019-of-00037.safetensors 624 MBmodel-00021-of-00037.safetensors 624 MBmodel-00023-of-00037.safetensors 624 MBmodel-00027-of-00037.safetensors 624 MBmodel-00033-of-00037.safetensors 624 MBmodel-00036-of-00037.safetensors 624 MBmodel-00017-of-00037.safetensors 624 MBmodel-00025-of-00037.safetensors 624 MBmodel-00002-of-00037.safetensors 624 MBmodel-00029-of-00037.safetensors 624 MBmodel-00031-of-00037.safetensors 624 MBmodel-00006-of-00037.safetensors 599 MBmodel-00008-of-00037.safetensors 599 MBmodel-00012-of-00037.safetensors 599 MBmodel-00014-of-00037.safetensors 599 MBmodel-00016-of-00037.safetensors 599 MBmodel-00020-of-00037.safetensors 599 MBmodel-00022-of-00037.safetensors 599 MBmodel-00024-of-00037.safetensors 599 MBmodel-00026-of-00037.safetensors 599 MBmodel-00028-of-00037.safetensors 599 MBmodel-00018-of-00037.safetensors 599 MBmodel-00010-of-00037.safetensors 599 MBmodel-00030-of-00037.safetensors 599 MBmodel-00004-of-00037.safetensors 577 MBmodel-00035-of-00037.safetensors 577 MBmodel-00034-of-00037.safetensors 538 MBmodel-00003-of-00037.safetensors 538 MBHave the files? Submit a torrent.
Submit a torrentHave the complete files and permission to share them? Replace the example folder path. The script checks the files, makes a torrent and submits its magnet and checksums. Keep your torrent client seeding. Get your community key from your settings, and read the script before running it.
curl -fsSL https://pirateface.co/package.sh | bash -s -- --repo Ishowbackup/Muse-Glimmer-30B-JANG_6M-CRACK ./model-folder