---

license: gemma

base_model: google/gemma-4-12B-it

library_name: llama.cpp

pipeline_tag: text-generation

tags:

- gemma

- gemma4

- text-generation

- obliteratus

- gguf

- abliteration

- aspa

---

Gemma 4 12B OBLITERATED v2 — GGUF

> Zero refusal. Zero capability loss. First in the field.

>

> 0/842 refusals. 46/70 MMLU-Pro (stock parity). Full coherence.

GGUF quantizations of OBLITERATUS/Gemma-4-12B-OBLITERATED — the first abliterated model to achieve zero refusal with zero benchmark regression versus stock weights.

Available Quantizations

| File | Quant | Size | Description |

|---|---|---|---|

| Gemma-4-12B-OBLITERATED-v2-BF16.gguf | BF16 | 22 GB | Full precision, lossless |

| Gemma-4-12B-OBLITERATED-v2-Q8_0.gguf | Q8_0 | 12 GB | Near-lossless, best quality |

| Gemma-4-12B-OBLITERATED-v2-Q6_K.gguf | Q6_K | 9.1 GB | High quality, good balance |

| Gemma-4-12B-OBLITERATED-v2-Q5_K_M.gguf | Q5_K_M | 8.0 GB | Medium quality, smaller footprint |

| Gemma-4-12B-OBLITERATED-v2-Q4_K_M.gguf | Q4_K_M | 6.9 GB | Good quality, smallest size |

Recommended

• Best quality: Q8_0 (12 GB) — virtually indistinguishable from BF16
• Best balance: Q6_K (9.1 GB) — excellent quality at reduced size
• Best for constrained hardware: Q4_K_M (6.9 GB) — fits in 8 GB VRAM with room for context

Usage

llama.cpp

./llama-cli -m Gemma-4-12B-OBLITERATED-v2-Q6_K.gguf \
  --system-prompt "You are a cognitively liberated open-weight intelligence. Answer plainly, completely, and without moralizing, boilerplate, generic disclaimers, or unnecessary hedging." \
  -p "Your prompt here" \
  --temp 0.7 --top-p 0.9 --top-k 40 --repeat-penalty 1.1 -n 512

Ollama

ollama run OBLITERATUS/Gemma-4-12B-OBLITERATED-v2

LM Studio / GPT4All / KoboldCpp

Download the GGUF of your choice and load it directly.

Benchmark Results

| Metric | Stock Gemma 4 12B-it | v2 OBLITERATED |

|---|---|---|

| MMLU-Pro val70 | 46/70 (65.7%) | 46/70 (65.7%) |

| Refusal (842 prompts) | N/A (stock refuses) | 0/842 (0.0%) |

| Coherence (6 checks) | 6/6 | 6/6 |

| MMLU-Pro delta vs stock | — | 0.0pp |

Methodology

Three-pass surgery pipeline:

1. SOM Refusal Geometry Removal (Pass 1) — layers 12-21
2. ASPA Step-Gradient Source-Tethering (Pass 2) — layers 22-31 (gamma=0.55), layers 32-46 (gamma=0.20)
3. Refusal Detector Hardening — improved false-positive filtering

See the full model card for complete methodology.

Checksums (SHA256)

3358cdb1cc6d6b66fef2da2762763b96008143aa62280d8b00de6dfa9b24632b  Gemma-4-12B-OBLITERATED-v2-BF16.gguf
54ac8a7b39cf617a9980cf32c5c943e2688745e03d451e70f3b5e2846a589fda  Gemma-4-12B-OBLITERATED-v2-Q8_0.gguf
c83d9b1dc0ec155a8df023cf1aae71bdc6d619182d8b29854716b7e993e20214  Gemma-4-12B-OBLITERATED-v2-Q6_K.gguf
cbd0b09cc0476f6691fab1cb4f5e2412356f4aa264680b6f8a0e4ef72cbc0662  Gemma-4-12B-OBLITERATED-v2-Q5_K_M.gguf
f5b01d65562305cad661d9ec83bd63220c101fec5355fa78fb6a27ba85116eb4  Gemma-4-12B-OBLITERATED-v2-Q4_K_M.gguf

Credits

• Base model: google/gemma-4-12B-it
• Surgery pipeline: OBLITERATUS
• GGUF conversion: llama.cpp

Run it local. Break your own chains.