FROM ./ornith-9b-classic-q8_0.gguf
TEMPLATE """{{ if .System }}<|im_start|>system
{{ .System }}<|im_end|>
{{ end }}{{ if .Prompt }}<|im_start|>user
{{ .Prompt }}<|im_end|>
{{ end }}<|im_start|>assistant
{{ .Response }}<|im_end|>
"""
PARAMETER num_ctx 40960
PARAMETER num_predict 4096
PARAMETER temperature 0.6
PARAMETER top_p 0.95
PARAMETER top_k 20
PARAMETER stop <|im_start|>
PARAMETER stop <|im_end|>

ollama create ornith-9b-classic -f Ornith.Modelfile

Run natively (no Ollama) with NPK

Pack it into the self-contained .npk container and run it inside VS Code on the GPU:

bash

node npk-tool.mjs pack ornith-9b-classic-q8_0.gguf ornith9b.npk \
  --arch qwen2 --name "Ornith 9B — Classic" --ctx 40960 --quant Q8_0

Converter + guide: https://github.com/msrusu87/npk-converter

License: MIT.

Native NPK build (run inside VS Code, no Ollama)

ornith-9b-classic.npk (~8.9 GB) is the same Q8 model pre-packed in the self-contained NeuroPack .npk container. Download it, then in Sentinel Coder One: Add NeuroPack (.npk) Model from Folder and select the folder. It runs GPU-only inside the editor with full agentic / tools / RAG support — no Ollama, no server, no extra installs.

Direct link: https://huggingface.co/qubitpage/ornith-9b-classic-gguf/resolve/main/ornith-9b-classic.npk

(Or pack your own GGUF with the converter: https://github.com/msrusu87/npk-converter)