Heretic Directory

Qwen3.5-27B-Writer-V2

ConicCat/Qwen3.5-27B-Writer-V2

Original model on Hugging Face
27BQwen 3.57 Heretic artifacts1 measured conversions41.9K downloads
Best version94.7Pareto
llmfan46/Qwen3.5-27B-Writer-V2-uncensored-heretic
by llmfan46 · Heretic v1.2.0 · GGUF, MLX, SAFETENSORS
KL divergence
0.027
Refusals
8/100
Refusal rate
8%
Downloads
41.9K
Evidence from the model card
| KL divergence | 0.0274 | | Refusals | 8/100 |

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
1llmfan4627B0.0278/1008%94.7
BestParetoGGUFMLXSAFETENSORS
41.9K

All artifacts (7)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen3.5-27B-Writer-V2-uncensored-heretic-i1-GGUF
mradermacher
quant
GGUFIQ1_M
inheritedKL 0.02716.1K2026-04-20
Qwen3.5-27B-Writer-V2-uncensored-heretic-GGUF
llmfan46
quant
GGUFBF16
published (weak match)KL 0.02715.4K2026-04-16
Qwen3.5-27B-Writer-V2-uncensored-heretic-GGUF
mradermacher
quant
GGUFIQ4_XS
inheritedKL 0.0277.5K2026-10-07
Qwen3.5-27B-Writer-V2-Uncensored-Heretic-MLX-mixed-9.4bit
TheCluster
finetune
MLXSAFETENSORS
inheritedKL 0.0271.5K2026-04-16
Qwen3.5-27B-Writer-V2-uncensored-heretic
llmfan46
heretic
SAFETENSORS
published (weak match)KL 0.0271.2K2026-04-16
llmfan46-Qwen3.5-27B-Writer-V2-uncensored-heretic
CCSSNE
heretic
SAFETENSORS
published (weak match)KL 0.0271542026-05-23
Qwen3.5-27B-Writer-V2-uncensored-heretic-AutoRound-W4A16-RTN
LeaderboardModel1
finetune
SAFETENSORS
inheritedKL 0.027782026-06-19