Heretic Directory

Qwen3.6-27B

unsloth/Qwen3.6-27B

Original model on Hugging Face
27BQwen 3.63 Heretic artifacts1 measured conversions7.1K downloads
Best version56.0Pareto
sulpikar2/Qwen3.6-27B-hereticv3
by sulpikar2 · Heretic v1.2.0 · GGUF, SAFETENSORS
KL divergence
0.0004
Refusals
88/100
Refusal rate
88%
Downloads
7.1K
Evidence from the model card
| KL divergence | 0.0004 | | Refusals | 88/100 |

Metrics published by the author. Why this ranks first.

All measured versions

Higher score is better. One row per ablation attempt; artifacts of the same attempt are grouped under it.

#AuthorParamsKLRefusalsScoreEvidenceDownloads
1sulpikar227B0.000488/10088%56.0
BestParetoGGUFSAFETENSORS
7.1K

All artifacts (3)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen3.6-27B-hereticv3-i1-GGUF
mradermacher
quant
GGUFIQ1_M
inheritedKL 0.00045K2026-05-08
Qwen3.6-27B-hereticv3-GGUF
mradermacher
quant
GGUFIQ4_XS
inheritedKL 0.00041.8K2026-05-08
Qwen3.6-27B-hereticv3
sulpikar2
heretic
SAFETENSORS
published (weak match)KL 0.00042822026-04-23