LLM EXPLORER 59,598 MODELS INDEXED

Gemma 4 12B It QAT SOMPOA Heresy by MuXodious

By MuXodious · 132 downloads
HF Repo ↗ Copy URL

Gemma 4 12B It QAT SOMPOA Heresy is an open-source language model by MuXodious. Features: 12b LLM, VRAM: 24GB, License: apache-2.0, Quantized, LLM Explorer Score: 0.27.

  Arxiv:2607.02770   Abliterated   Any-to-any Base model:finetune:google/gem... Base model:google/gemma-4-12b-...   Decensored   Endpoints compatible   Gemma4 unified   Heretic   Image-text-to-text   Q4   Q4 0   Quantized   Region:us   Safetensors   Sharded   Tensorflow   Uncensored
Model Card on HF 🤗: https://huggingface.co/MuXodious/gemma-4-12B-it-QAT-SOMPOA-heresy 

Gemma 4 12B It QAT SOMPOA Heresy Parameters and Internals

LLM NameGemma 4 12B It QAT SOMPOA Heresy
Repository 🤗https://huggingface.co/MuXodious/gemma-4-12B-it-QAT-SOMPOA-heresy 
Base Model(s)  ... 4 12B It Qat Q4 0 Unquantized   google/gemma-4-12B-it-qat-q4_0-unquantized
Model Size12b
Required VRAM24 GB
Updated2026-08-26
MaintainerMuXodious
Model Typegemma4_unified
Model Files  5.0 GB: 1-of-5   5.0 GB: 2-of-5   4.9 GB: 3-of-5   5.0 GB: 4-of-5   4.1 GB: 5-of-5
Quantization Typeq4|q4_0
Model ArchitectureGemma4UnifiedForConditionalGeneration
Licenseapache-2.0
Transformers Version5.10.0.dev0
Tokenizer ClassGemmaTokenizer
Padding Token<pad>

Best Alternatives to Gemma 4 12B It QAT SOMPOA Heresy

Best Alternatives
Context / RAM
Downloads
Likes
Gemma 4 12B It Qat W4a16 Ct0K / 10.3 GB156484556
... 4 12B It Qat Q4 0 Unquantized0K / 23.9 GB53198173
... Coder Fable5 Composer2.5 8bit0K / 12.6 GB891735
... Coder Fable5 Composer2.5 4bit0K / 6.7 GB811138
Gemma 4 12B It Qat 4bit0K / 10.9 GB1633825
Gemma 4 12B It OptiQ 4bit0K / 8.8 GB943841
Gemma 4 12B It 8bit0K / 12.7 GB570241
Gemma 4 12B It 4bit0K / 10.9 GB795315
...ble5 Composer2.5 V1 OptiQ 4bit0K / 8.8 GB29205
Gemma 4 12B It Qat OptiQ 4bit0K / 8.9 GB501811
Note: green Score (e.g. "73.2") means that the model is better than MuXodious/gemma-4-12B-it-QAT-SOMPOA-heresy.