license: apache-2.0 base_model: Jiunsong/SuperGemma-4-12b-abliterated base_model_relation: quantized library_name: gguf pipeline_tag: text-generation tags:
- gemma4
- abliterated
- gguf
- q4_k_m
- 4-bit
- quantized
- llama.cpp language:
- en
- ko
SuperGemma-4-12b-abliterated - GGUF 4-bit
This repository contains a llama.cpp GGUF Q4_K_M quantization of the BF16 parent model.
Parent model: Jiunsong/SuperGemma-4-12b-abliterated
Parent Benchmark Summary
| Benchmark | Gemma4 12B original | SuperGemma-4-12b-abliterated | Delta |
|---|---|---|---|
| Overall public top-5 500 | 23.8 | 44.6 | +20.8 |
| GPQA Diamond | 10.0 | 19.0 | +9.0 |
| MMLU-Pro | 17.0 | 18.0 | +1.0 |
| IFEval | 61.0 | 59.0 | -2.0 |
| HumanEval+ | 18.0 | 46.0 | +28.0 |
| MBPP+ | 13.0 | 81.0 | +68.0 |
Usage
llama-cli \
-m SuperGemma-4-12b-abliterated-Q4_K_M.gguf \
--single-turn \
--reasoning off \
--no-display-prompt \
-p "Write a short release note."
Notes
- This quantized build is provided for smaller local inference deployments.
- For llama.cpp CLI smoke tests, use
--single-turn --reasoning offfor clean final-answer output. - For benchmark numbers and validation details, see the BF16 parent repository.