Heretic Directory

Llama-3.3-8B-Instruct

allura-forge/Llama-3.3-8B-Instruct

Original model on Hugging Face
8BLlama 322 Heretic artifacts1 measured conversions832.1K downloads
Best version86.5Pareto
Babsie/DavidAU_L3.3_8B_OpusR_HUC
by Babsie · SAFETENSORS
KL divergence
0.150
Refusals
14/100
Refusal rate
14%
Downloads
147
Evidence from the model card
(Details: org censorship: 89/100 ; Heretic version: Refusals: 14/100, KL divergence: 0.1498 (less than 1 is excellent, 0=perfect) )

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
1Babsie8B0.15014/10014%86.5
BestParetoSAFETENSORS
147

All artifacts (22)

RepositoryTypeFormatsMetricsDownloadsUpdated
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-i1-GGUF
mradermacher
quant
GGUFIQ1_M
none744.7K2026-01-02
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-8Bit
coderavi
quant
MLXSAFETENSORS
none23.6K2026-02-01
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-GGUF
mradermacher
quant
GGUFIQ4_XS
none14.4K2026-01-02
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-i1-GGUF
lingfengyaovo
quant
GGUFIQ1_M
none8.2K2026-04-24
DavidAU_L3.3_8B_OpusR_HUC-i1-GGUF
mradermacher
quant
GGUFIQ1_M
none7.4K2026-03-07
Llama3.3-8B-Instruct-Thinking
wiyamop
quant
GGUFIQ1_M
none4.5K2026-02-19
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-i1-GGUF
leaderaigmbh
quant
GGUFIQ1_M
none4.2K2026-03-26
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-Q8_0-GGUF
speedstarrr
quant
GGUFQ8_0
none4K2026-01-11
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-i1-GGUF
Mal1cc
quant
GGUFIQ1_M
none3.9K2026-04-20
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-8Bit
vitalii39
quant
MLXSAFETENSORS
none3.1K2026-05-25
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-IQ4_NL-GGUF
SpecialPatrolGroup
quant
GGUFIQ4_NL
none2.8K2026-02-07
New-Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-8Bit
Unitedp2p
quant
MLXSAFETENSORS
none2.2K2026-03-19
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-2Bit
alexgusevski
quant
MLXSAFETENSORS
none1.9K2026-01-12
DavidAU_L3.3_8B_OpusR_HUC-GGUF
mradermacher
quant
GGUFIQ4_XS
none1.8K2026-03-07
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-fp16
alexgusevski
quant
MLXSAFETENSORS
none1.2K2026-01-12
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-modified
HuyEdits
quant
GGUFSAFETENSORS
none8972026-03-01
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-4Bit
alexgusevski
quant
MLXSAFETENSORS
none7612026-01-12
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-8Bit
alexgusevski
quant
MLXSAFETENSORS
none6792026-01-12
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-3Bit
alexgusevski
quant
MLXSAFETENSORS
none6382026-01-12
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-6Bit
alexgusevski
quant
MLXSAFETENSORS
none6002026-01-12
Llama3.3-8B-Instruct-Thinking-Heretic-Uncensored-Claude-4.5-Opus-High-Reasoning-mlx-5Bit
alexgusevski
quant
MLXSAFETENSORS
none5622026-01-12
DavidAU_L3.3_8B_OpusR_HUC
Babsie
finetune
SAFETENSORS
published (weak match)KL 0.1501472026-03-05

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