Ornith-1.0-35B-MXFP4-MOE-GGUF

Ornith 1.0 35B — MXFP4 MoE GGUF

MXFP4 MoE quantization of deepreinforce-ai/Ornith-1.0-35B, a 35B parameter Qwen3.5 MoE coding agent with 256 experts (8 active per token).

About the Model

Ornith-1.0-35B is the lightweight member of the Ornith family, designed for efficient single-GPU deployment.

• State-of-the-Art Coding Agents: Post-trained on top of Qwen 3.5, achieving state-of-the-art performance among open-source models
• Self-Improving Training Framework: Ornith-1.0 employs RL to learn to generate not only solution rollouts, but also the scaffold that drives those rollouts
• 35B total parameters with 8B active per token (256 experts, 8 active)
• 40-layer MoE architecture with sliding + full attention hybrid
• 262K context window
• MIT License — globally accessible, no regional limitations

Architecture

• Text model: Qwen3.5 MoE — 40 layers, 2048 hidden, 256 experts (8 active/token)
• Vocabulary: 248,320 tokens

Quantization

Quantized from the BF16 safetensors using llama.cpp (build 537).

MXFP4 MoE uses microscaling FP4 for expert weights and Q8_0 for non-expert tensors, optimized for MoE architectures.

Files

Usage

llama-server \
  -m ornith-1.0-35b-mxfp4_moe.gguf \
  -ngl 99 \
  --host 0.0.0.0 \
  --port 8080

Hardware Requirements

• Minimum: 20 GB VRAM
• Recommended: 24+ GB VRAM for full GPU offload

License

MIT