Ornith-1.5-35B-A3B-Abliterated is created through directional ablation surgery across all 40 hybrid layers, orthogonally projecting refusal directions out from attention and MLP down-projection matrices.

This dedicated repository provides optimized Dynamic GGUF quantizations for local inference in Ollama, LM Studio, and Llama.cpp.

Quantization Details

File Name	Format	Size	Description
`Ornith-1.5-35B-Abliterated-Dynamic-Q4_K_M.gguf`	Q4_K_M (Dynamic)	~19.71 GB	Balanced performance for 12GB - 16GB VRAM GPUs.
`Ornith-1.5-35B-Abliterated-Dynamic-Q5_K_M.gguf`	Q5_K_M (Dynamic)	~23.03 GB	Higher precision retention.
`Ornith-1.5-35B-Abliterated-Dynamic-Q3_K_M.gguf`	Q3_K_M (Dynamic)	~15.61 GB	Lower memory footprint.
`Ornith-1.5-35B-Abliterated-Q8_0.gguf`	Q8_0	~34.37 GB	High precision reference quantization.

How to Run

1. Using Ollama

ollama run hf.co/alztrk/Ornith-1.5-35B-A3B-Abliterated-GGUF:Ornith-1.5-35B-Abliterated-Dynamic-Q4_K_M.gguf

2. Using LM Studio / Llama.cpp

llama-cli.exe -hf alztrk/Ornith-1.5-35B-A3B-Abliterated-GGUF -hff Ornith-1.5-35B-Abliterated-Dynamic-Q4_K_M.gguf -p "<|im_start|>user\nHello!<|im_end|>\n<|im_start|>assistant\n" -ngl 24 -c 8192

License & Attribution

Derived from ornith-ai/Ornith-1.5-35B-A3B, licensed under the Apache 2.0 / MIT License.