DarkArtsForge/huihui-ai_Huihui-GLM-4.5-Air-abliterated-Q3_K_M-GGUF

110B • Updated 29 days ago • 244

Note Still the best model for local dataset generation. If you have more than 64GB RAM you can use Q4_K_M from concedo/Huihui-GLM-4.5-Air-abliterated-lossytensors. This model is far better than Gemma 4 26BA4B in all my tests.

mlabonne/gemma-3-27b-it-abliterated

Image-Text-to-Text • 27B • Updated Mar 21, 2025 • 2.87k • • 331

Note Best for complex system prompts. Better than all G4 models in my testing. Similar quality to GLM 4.5 Air but slower.

TheDrummer/Fallen-Command-A-111B-v1

111B • Updated Mar 24, 2025 • 12 • 26

Note Recommended IQ2_XXS for offloading. Insert the output of this model into the < think > tags of newer models to create "fallen" versions of them.

SicariusSicariiStuff/Assistant_Pepe_8B

266k • Updated May 3 • 247 • 74

Note 8B is more stable than 32B or 70B on my system. Great for producing unhinged output.

DarkArtsForge/MN-Raven-12B-v1

Text Generation • Updated 16 days ago • 145 • • 8

Note Excellent for producing 19th century style prose.

DarkArtsForge/Morax-24B-v2

Text Generation • Updated 25 days ago • 204 • 6

Note Extremely morbid and deranged.

EldritchLabs/MN-12B-Mag-Mell-R1-Uncensored-Scale1.2

Text Generation • 12B • Updated Mar 5 • 150 • • 8

Note Best for when all else fails

TheDrummer/Tiger-Gemma-9B-v3

9B • Updated Oct 4, 2024 • 12 • 61

Note Mostly obsolete now except for low VRAM agentic use. Better than Gemma 3 12B in my tests. IQ3_S fits entirely on 8GB VRAM. Tiger 9B works surprisingly well for system prompts. It was the best budget model of its time and is still my favorite 9B along with Delirium by sam paech.

Naphula/DarkForest-20B-v2.0-Erebus-Edition

20B • Updated Nov 23, 2025 • 9 • 12

Note Creative and uncensored RP, dark and unhinged, the OG of this list. Powerful datasets on a limited engine.