Ornith 35B quantized down to UD-IQ3_XXS following the settings used by unsloth for their Qwen3.6-35B-A3B model, and with its MTP head grafted on. Jinja Exception: System message must be at the beginning exception removed from chat template.

fits in 16GB of VRAM with 200k of q5_0/q5_0 context
MTP draft=2 and 8k of turbo4 context averages ~50% draft acceptance on longer prompts
on an RTX 5060 Ti, can peak at around 120tg/s for short prompts, and maintains 70+tg/s even at 60k tokens

Citation

Original Model: <a href="https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">https://huggingface.co/deepreinforce-ai/Ornith-1.0-35B
Unsloth Model: <a href="https://huggingface.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">https://huggingface.co/unsloth/Qwen3.6-35B-A3B-MTP-GGUF
Ornith Blog: <a href="https://deep-reinforce.com/ornith.html" target="_blank" rel="noopener noreferrer" class="text-blue-400 hover:text-blue-300 underline underline-offset-2 transition-colors">https://deep-reinforce.com/ornith.html

bibtex

@misc{ornith-35b,
    title = {{Ornith-1.0-35B}: Agentic Coding, Open to All},
    url = {https://deep-reinforce.com/ornith_1_0.html},
    author = {{DeepReinforce Team}},
    year = {2026}
}

zacaj/Ornith-1.0-35B-unsloth_UD-IQ3_XXS-MTP-gguf

Links

Citation

zacaj/
Ornith-1.0-35B-unsloth_UD-IQ3_XXS-MTP-gguf