DeepReinforce Releases Self-Planning Local LLM Ornith-1.0

DeepReinforce released Ornith-1.0, an open-source agentic coding LLM that runs locally via Ollama, in late June 2026. Its defining feature is "self-scaffolding": the model drafts its own plan before coding and then grades both the strategy and the output. Announced on the team's official blog and Hugging Face, it ships in four sizes — 9B and 31B Dense, plus 35B and 397B MoE — under an MIT license. It runs immediately with ollama run ornith (or ornith:9b / ornith:35b), and GGUF builds are distributed via Hugging Face for use through llama.cpp or Ollama. It operates fully local and offline with no API key, supporting everything from edge devices to high-end GPUs.