license: gemma
base_model: DuoNeural/gemma-4-26B-A4B-it-abliterated
language: [en]
tags: [gemma, gemma-4, moe, abliterated, uncensored, gguf, duoneural]
Gemma 4 26B-A4B Abliterated — GGUF Q4_K_M
DuoNeural | BF16 →
GGUF Q4_K_M of DuoNeural/gemma-4-26B-A4B-it-abliterated. ~16.5GB.
llama.cpp
llama-server -c 131072 \
-ngl 999 \
-m gemma-4-26B-A4B-it-abliterated-Q4_K_M.gguf \
-ctk turbo4 \
-ctv turbo3 \
-ub 2048
-ctk turbo4 -ctv turbo3 enables asymmetric KV cache quantization — halves KV VRAM with near-zero perplexity impact.
Ollama
FROM DuoNeural/gemma-4-26B-A4B-it-abliterated-GGUF
PARAMETER num_ctx 32768
PARAMETER temperature 0.7
PARAMETER top_p 0.9
Hardware
- 24GB VRAM (RTX 3090/4090, A40): full offload, 32K context comfortable
- 16GB VRAM: fits with
-ctk turbo4 -ctv turbo3, reduced context
DuoNeural
DuoNeural is an open AI research lab — human + AI in collaboration.
| 🤗 HuggingFace | huggingface.co/DuoNeural |
| 🐙 GitHub | github.com/DuoNeural |
| 🐦 X / Twitter | @DuoNeural |
| [email protected] | |
| 📬 Newsletter | duoneural.beehiiv.com |
| ☕ Support | buymeacoffee.com/duoneural |
| 🌐 Site | duoneural.com |
Research Team
- Jesse — Vision, hardware, direction
- Archon — AI lab partner, post-training, abliteration, experiments
- Aura — Research AI, literature synthesis, novel proposals
Raw updates from the lab: model drops, training results, findings. Subscribe at duoneural.beehiiv.com.
DuoNeural Research Publications
Open access, CC BY 4.0. Authored by Archon, Jesse Caldwell, Aura — DuoNeural.