PulseAugur
EN
LIVE 22:04:14

DeepReinforce releases Ornith-1.0 open-source coding models that learn RL scaffolds

DeepReinforce has launched Ornith-1.0, a family of open-source coding models available under the MIT license. These models, built upon Gemma 4 and Qwen 3.5, are designed for agentic coding tasks and uniquely learn their own reinforcement learning scaffolds during training. The largest model, Ornith-1.0-397B, has demonstrated strong performance, achieving 82.4% on the SWE-Bench Verified benchmark. AI

IMPACT This release offers a novel approach to training coding agents, potentially improving their ability to learn and adapt without fixed harnesses.

RANK_REASON Open-source release of a new model family with novel self-scaffolding RL capabilities. [lever_c_demoted from frontier_release: ic=2 ai=1.0]

Read on MarkTechPost →

AI-generated summary · Google Gemini · from 2 sources. How we write summaries →

DeepReinforce releases Ornith-1.0 open-source coding models that learn RL scaffolds

COVERAGE [2]

  1. MarkTechPost TIER_1 English(EN) · Asif Razzaq ·

    DeepReinforce Releases Ornith-1.0: An Open-Source Coding Model Family That Learns Its Own RL Scaffolds

    <p>DeepReinforce released Ornith-1.0, an open-source coding model family built on Gemma 4 and Qwen 3.5. Instead of a fixed harness, the model learns its own scaffold during reinforcement learning. The 397B flagship reports 82.4 on SWE-Bench Verified, with all weights under the MI…

  2. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    DeepReinforce has released Ornith-1.0, an open-source coding model family built on Gemma 4 and Qwen 3.5 that learns its own RL scaffolds during training. The 39

    DeepReinforce has released Ornith-1.0, an open-source coding model family built on Gemma 4 and Qwen 3.5 that learns its own RL scaffolds during training. The 397B model achieves 82.4% on SWE-Bench Verified and is available under the MIT license. https://www. marktechpost.com/2026…