This page may contain stale information. Last updated: 2026-06-28
Overview
DeepReinforce is an AI research team specializing in reinforcement learning for agentic coding. Prior open-source work includes CUDA-L1 and the IterX code-agent optimization loop.
Recent Developments
- 2026-06-25: Released ornith-1 — four MIT-licensed coding models (9B–397B MoE) with self-scaffolding RL training (2026-06-26-deepreinforce-ornith-official-blog)
Related
- ornith-1
- open-source-ai
- reinforcement-learning
- ai-coding-tools
- gemma-4
- qwen
- deepreinforce-ornith-1-open-source-coding