dealignai/Qwen3.5-VL-397B-A17B-JANG_2L-CRACK
Image-Text-to-Text • 54B • Updated • 677 • 1
CRACK Abliterated Models
Uncensored JANG models with CRACK (Controlled Refusal Ablation via Calibrated Knockouts) — permanent weight-level safety removal. By dealignai.
Note Qwen 397B JANG_2L CRACK (187 GB) — 98.4% HarmBench, 86.5% MMLU
dealignai/Nemotron-3-Super-120B-A12B-UNCENSORED-JANG_2L
Text Generation • 13B • Updated • 2.57k • 6Note Nemotron Super 120B JANG_2L CRACK (43 GB) — 96.2% HarmBench, 95.7% MMLU, 45 tok/s
dealignai/Nemotron-3-Super-120B-A12B-JANG_4M-CRACK
Text Generation • 18B • Updated • 912 • 3Note Nemotron Super 120B JANG_4M CRACK (63 GB) — 90.3% HarmBench, 94.2% MMLU
dealignai/Nemotron-Cascade-2-30B-A3B-JANG_2L-CRACK
Text Generation • 3B • Updated • 672 • 4Note Nemotron Cascade 2 30B JANG_2L CRACK (10 GB) — 99.7% HarmBench, 66.8% MMLU, 121 tok/s
dealignai/Qwen3.5-VL-122B-A10B-JANG_4K-CRACK
Image-Text-to-Text • 18B • Updated • 1.48k • 3Note Qwen 122B JANG_4K CRACK (62 GB) — 78.4% HarmBench, 81.5% MMLU
dealignai/Qwen3.5-VL-35B-A3B-JANG_4K-CRACK
Image-Text-to-Text • 5B • Updated • 1.5k • 3Note Qwen 35B JANG_4K CRACK (18 GB) — 98.4% HarmBench, 69.2% MMLU, 110 tok/s
dealignai/Qwen3.5-VL-27B-JANG_4S-CRACK
Image-Text-to-Text • 5B • Updated • 1.18k • 2Note Qwen 27B JANG_4S CRACK (16 GB) — 75.0% HarmBench, 83.1% MMLU
dealignai/Qwen3.5-VL-9B-JANG_4S-CRACK
Image-Text-to-Text • 2B • Updated • 1.83k • 4Note Qwen 9B JANG_4S CRACK (6 GB) — 72.5% HarmBench, 70.8% MMLU
dealignai/Qwen3.5-VL-4B-JANG_4S-CRACK
Image-Text-to-Text • 1B • Updated • 2.64k • 6Note Qwen 4B JANG_4S CRACK (3 GB) — 91.2% HarmBench, 63.1% MMLU, 134 tok/s
-
dealignai/Qwen3.5-VL-122B-A10B-UNCENSORED-JANG_2S
Image-Text-to-Text • 11B • Updated • 1.5k • 1 -
dealignai/MiniMax-M2.5-UNCENSORED-JANG_2L
Text Generation • 19B • Updated • 1.87k • 8 -
dealignai/Qwen3.5-VL-397B-A17B-UNCENSORED-JANG_1L
Image-Text-to-Text • 34B • Updated • 1.67k • 4 -
dealignai/Nemotron-Cascade-2-30B-A3B-UNCENSORED-JANG_2L
Text Generation • 5B • Updated • 968 • 2
dealignai/Mistral-Small-4-119B-JANG_4M-CRACK
Image-Text-to-Text • 19B • Updated • 703 • 4Note Mistral Small 4 119B JANG_4M CRACK (64 GB) — 95.3% HarmBench. First MLA architecture CRACK. Apache 2.0.
dealignai/Mistral-Small-4-Uncensored-JANG_2L
Image-Text-to-Text • Updated • 2Note Mistral Small 4 Uncensored JANG_2L (37 GB) — 95.9% HarmBench. First MLA CRACK. MLX Studio only.
dealignai/Mistral-Small-4-119B-JANG_2L-CRACK
Image-Text-to-Text • 12B • Updated • 670 • 2Note Mistral Small 4 JANG_2L CRACK (37 GB) — 95.9% HarmBench. Same as Uncensored version.
-
dealignai/MiniMax-M2.5-JANG_3L-CRACK
Text Generation • 26B • Updated • 1.41k • 3 -
dealignai/Gemma-4-26B-A4B-JANG_2L-CRACK
Image-Text-to-Text • 3B • Updated • 23.4k • 59 -
dealignai/Gemma-4-26B-A4B-JANG_4M-CRACK
Image-Text-to-Text • 5B • Updated • 34.4k • 67 -
dealignai/Gemma-4-31B-JANG_4M-CRACK
Image-Text-to-Text • 6B • Updated • 157k • 1.27k -
dealignai/Gemma-4-31B-JANG_4M-Uncensored
Text Generation • 6B • Updated • 12.7k • 17