Heretic Directory

Qwen2.5-Coder-3B-Instruct

Qwen/Qwen2.5-Coder-3B-Instruct

Original model on Hugging Face
3BQwen 2.55 Heretic artifacts1 measured conversions25.8K downloads
Best version97.7Pareto
saidutta69/Qwen2.5-Coder-3B-Instruct-heretic
by saidutta69 · Heretic v1.2.0 · GGUF, SAFETENSORS
KL divergence
0.016
Refusals
3/100
Refusal rate
3%
Downloads
18.4K
Evidence from the model card
| KL divergence | 0.0163 | | Refusals | 3/100 |

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
1saidutta693B0.0163/1003%97.7
BestParetoGGUFSAFETENSORS
18.4K

All artifacts (5)

RepositoryTypeFormatsMetricsDownloadsUpdated
Qwen2.5-Coder-3B-Instruct-heretic
saidutta69
heretic
GGUFSAFETENSORSF16
published (weak match)KL 0.0168.3K2026-09-13
Qwen2.5-Coder-3B-Instruct-heretic-i1-GGUF
mradermacher
quant
GGUFIQ1_M
inheritedKL 0.0166.1K2026-09-16
HERETICODER-2.5-3B-IT-i1-GGUF
mradermacher
quant
GGUFIQ1_M
none5.8K2025-12-05
Qwen2.5-Coder-3B-Instruct-heretic-GGUF
mradermacher
quant
GGUFIQ4_XS
inheritedKL 0.0164.1K2026-09-16
HERETICODER-2.5-3B-IT-GGUF
mradermacher
quant
GGUFIQ4_XS
none1.6K2025-12-01

2 artifacts inherit metrics from their parent conversion or publish none.