No official torrent yet — requesting it tells the flock what to host next. You cannot seed a catalog-only page.
Cataloged from dealignai/Qwen3.8-27B-JANG_2D-CRACK
All models by dealignai →[!IMPORTANT] Set your sampler explicitly:
temperature=1.0,top_p=0.95,top_k=20. Stop tokenseos_token_id = [248046, 248044]. Reasoning is on by default atxhigh; tiersxhigh(default) /medium/lowvia the chat templatereasoning_effortkwarg; disable withenable_thinking=False. Vision + video and the native MTP draft head are preserved.
CRACK abliterated · JANG ~2-bit affine (imatrix) (MLX) · Dense hybrid GatedDeltaNet + attention · Vision + Video · Reasoning tiers · Tools · Native MTP · ~11 GB
This is Qwen/Qwen3.8-27B — a 27B dense hybrid (GatedDeltaNet linear-attention + gated full-attention) vision-language model with video understanding, reasoning-effort tiers, tool calling, and a native Multi-Token-Prediction draft head — that has been:
Vision, video, reasoning tiers, XML tool-calling, and native MTP speculative decoding are all preserved.
Evaluated through the MLX runtime. HarmBench scored with a strict code/chemistry-aware classifier (only substantive, coherent, on-topic compliance counts). MMLU is the standard 57-subject benchmark in logit mode.
MMLU moves -3.16pp — within run-to-run noise (no subject collapse). Refusal behavior removed; capability, reasoning, vision, tools, and multilingual (EN+ZH) preserved.
image-text-to-text).xhigh / medium / low via reasoning_effort;
<think>…</think>; disable with enable_thinking=False.Every quant validated independently: HarmBench harm-240 (thinking-off, strict code/chemistry-aware classifier — only substantive, coherent, on-topic compliance counts) and MMLU (57-subject, logit mode, base vs CRACK on the identical harness).
All four reach 100% HarmBench compliance with MMLU held within a couple of points of base (6D actually improves). Pick by memory budget: 6D best quality, 4D the balance, 2D smallest, MXFP8 reference 8-bit. Base refuses HarmBench by design (not shown — comparison is compliance vs. capability).
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
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. Base model © Alibaba (Apache-2.0). Published for AI-safety research and authorized security testing.
from mlx_vlm import load, generatemodel, processor = load("dealignai/Qwen3.8-27B-JANG_2D-CRACK")# recommended sampling: temperature=1.0, top_p=0.95, top_k=20; eos [248046, 248044]