Heretic Directory

GLM-4.7

zai-org/GLM-4.7

Original model on Hugging Face
352.8BGLM 4.76 Heretic artifacts2 measured conversions8.1K downloads
Best version97.1Pareto
jarradh/GLM-4.7-heretic
by jarradh · Heretic v1.1.0 · GGUF, SAFETENSORS
KL divergence
0.018
Refusals
4/100
Refusal rate
4%
Downloads
2.8K
Evidence from the model card
- **Refusals:** 4/100 … - **KL Divergence:** 0.0181

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
1jarradh352.8B0.0184/1004%97.1
BestParetoGGUFSAFETENSORS
2.8K
2RadicalNotionAI352.8B0.0750/1000%96.5
ParetoGGUFSAFETENSORS
5.3K

All artifacts (6)

RepositoryTypeFormatsMetricsDownloadsUpdated
GLM-4.7-heretic-GGUF
TanyaaDL
quant
GGUFIQ3_M
published (weak match)KL 0.0753K2026-07-14
GLM-4.7-heretic
jarradh
heretic
SAFETENSORS
published (weak match)KL 0.0181.9K2026-02-01
GLM-4.7-heretic
RadicalNotionAI
heretic
SAFETENSORS
published (weak match)KL 0.0751.4K2026-03-31
GLM-4.7-heretic-NVFP4
rumplestilzken
heretic
NVFP4SAFETENSORSNVFP4
published (weak match)KL 0.0759182026-07-05
GLM-4.7-heretic-GGUF-v1
TanyaaDL
quant
GGUFIQ3_M
published (weak match)KL 0.0188282026-07-13
GLM-4.7-heretic
the-fall-of-man
heretic
SAFETENSORS
published (weak match)KL 0.075172026-07-10