Heretic Directory

Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled

Jackrong/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled

Original model on Hugging Face
27BQwen 3.51 Heretic artifacts1 measured conversions685 downloads
Best version89.0Pareto
groxaxo/Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-heretic-EXL3-6bpw
by groxaxo · Heretic v1.2.0 · EXL, GGUF, SAFETENSORS
KL divergence
0.0092
Refusals
21/100
Refusal rate
21%
Downloads
685
Evidence from the model card
| KL divergence | 0.0092 | | Refusals | 21/100 |

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
1groxaxo27B0.009221/10021%89.0
BestParetoEXLGGUFSAFETENSORS
685

All artifacts (1)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen3.5-27B-Claude-4.6-Opus-Reasoning-Distilled-heretic-EXL3-6bpw
groxaxo
quant
EXLSAFETENSORS
published (weak match)KL 0.00926852026-08-22