Ornith-1.0 is the June 2026 debut release from DeepReinforce: an open-weights (MIT) family of agentic coding models (9B dense through 397B MoE variants) built on Gemma 4 and Qwen 3.5 bases. Its core innovation is a self-improving training loop where the model jointly learns to solve tasks and to write the task-specific execution harnesses that guide them, enabling inspectable, purpose-built scaffolding without human-designed templates.

Read the official DeepReinforce launch page

First builds incoming — this lane fills in automatically as people start building with Ornith 1.0. Know one? Submit the URL.