Ornith AI's Ornith 1.5 models range from 9B to 397B with GGUF
Three new Ornith 1.5 models on Hugging Face include a 397B MoE and GGUF weights
Deep Dive
A Reddit user shared links to Hugging Face pages for Ornith-1.5-9B, Ornith-1.5-35B-A3B, Ornith-1.5-397B, plus GGUF versions of each, and asked if anyone has tried them yet. The user added a disclaimer that they're not affiliated with Ornith—they just browse Hugging Face for new models every 30 minutes or so.
Key Points
- Ornith AI released three Ornith 1.5 models: 9B, 35B-A3B, and 397B on Hugging Face
- The 35B-A3B uses Mixture-of-Experts with only 3B active parameters for efficient inference
- GGUF quantized versions are available for all sizes, enabling local deployment via llama.cpp
Why It Matters
Open-weight models up to 397B give developers flexible, self-hosted alternatives to proprietary AI for cost and privacy.