Heretic Directory

Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled

hesamation/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled

Original model on Hugging Face
35B · 3B activeQwen 3.63 Heretic artifacts1 measured conversions8.6K downloads
Best version94.4Pareto
ghecko78/Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-heretic
by ghecko78 · Heretic v1.2.0 · GGUF, SAFETENSORS
KL divergence
0.066
Refusals
5/100
Refusal rate
5%
Downloads
8.6K
Evidence from the model card
| KL divergence | 0.0660 | | Refusals | 5/100 |

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
1ghecko7835B · 3B active0.0665/1005%94.4
BestParetoGGUFSAFETENSORS
8.6K

All artifacts (3)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-heretic-GGUF
ghecko78
quant
GGUFQ4_K_M
inheritedKL 0.0668K2026-05-12
Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-heretic
ghecko78
heretic
SAFETENSORS
published (weak match)KL 0.0666122026-04-26
Qwen3.6-35B-A3B-Claude-4.6-Opus-Reasoning-Distilled-heretic
modeler4d
heretic
SAFETENSORS
published (weak match)KL 0.066292026-09-29