Ornith-1.0-9B and Ornith-1.0-35B models now available on Amazon SageMaker JumpStart

DeepReinforce AI has made its Ornith-1.0-9B and Ornith-1.0-35B models available through Amazon SageMaker JumpStart, adding two new foundation models to AWS’s catalog. The 9B-parameter model is designed for efficient single-GPU deployment, featuring a 256K-token context window and 32 decoder layers, making it suitable for individual developers and small teams. It is post-trained on Qwen 3.5 9B and released under the MIT license, emphasizing accessibility and performance on consumer-grade hardware.

The Ornith-1.0-35B model is a 35B-parameter Mixture-of-Experts (MoE) variant post-trained on Qwen 3.5, offering stronger agentic coding capabilities while remaining deployable on a single 24GB consumer GPU. Its MoE architecture activates only a subset of parameters per token, balancing quality and efficiency for production workloads. Both models employ a self-improving training framework using reinforcement learning to optimize scaffold generation and solution rollouts, enhancing code quality and search trajectories.

These models achieve state-of-the-art performance among open-source models of comparable size on coding benchmarks such as Terminal-Bench 2.1, SWE-Bench, NL2Repo, and OpenClaw. The lightweight 9B model is tailored for efficient single-GPU deployment, while the 35B MoE model delivers higher performance without requiring specialized hardware. Both are part of the open-source Ornith 1.0 family, released under the MIT license for broad accessibility.

Customers can deploy either model with minimal setup via Amazon SageMaker JumpStart, either through the SageMaker console or the SageMaker Python SDK. The integration allows users to address specific AI use cases with just a few clicks, leveraging the models’ agentic coding capabilities. For deployment guidance, users can refer to the Amazon SageMaker JumpStart documentation for step-by-step instructions and best practices.