unsloth/Ornith-1.0-9B

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Jul 17, 2026License:mitArchitecture:Transformer0.0K Open Weights Featherless Exclusive Cold

Ornith-1.0-9B by unsloth is a 9 billion parameter model from the Ornith-1.0 family, specifically designed for agentic coding tasks. It utilizes a self-improving training framework based on Reinforcement Learning to optimize solution rollouts and scaffolds. This model achieves state-of-the-art performance among open-source models of comparable size on coding benchmarks like Terminal-Bench 2.1, SWE-Bench, NL2Repo, and OpenClaw, making it highly effective for agentic coding and tool-calling applications.

Loading preview...

Ornith-1.0-9B: Agentic Coding Model

Ornith-1.0-9B, developed by unsloth, is a 9 billion parameter model optimized for agentic coding. It is part of the Ornith-1.0 family, which employs a unique self-improving training framework using Reinforcement Learning. This framework enables the model to generate not only solution rollouts but also the underlying scaffolds, leading to better search trajectories and higher-quality solutions.

Key Capabilities

  • State-of-the-Art Agentic Coding: Achieves leading performance on coding benchmarks such as Terminal-Bench 2.1, SWE-Bench, NL2Repo, and OpenClaw, outperforming other models in its size class.
  • Self-Improving Training: Leverages RL to jointly optimize solution rollouts and scaffolds, enhancing problem-solving capabilities.
  • Efficient Deployment: As the most lightweight member of the Ornith family, it is designed for efficient single-GPU deployment.
  • Reasoning and Tool-Calling: Functions as a reasoning model, providing chain-of-thought in a separate reasoning_content field and surfacing tool calls in an OpenAI-compatible format.
  • MIT Licensed: Globally accessible and free from regional restrictions.

Good For

  • Agentic Coding: Ideal for tasks requiring autonomous code generation, debugging, and problem-solving within agent frameworks.
  • Terminal-Based Coding Agents: Optimized for integration with coding CLIs to understand large codebases and automate development workflows.
  • Tool-Calling Applications: Excels at emitting well-formed function calls, making it suitable for applications requiring interaction with external tools and APIs.
  • Local Inference and Fine-tuning: Can be loaded for fast local inference or fine-tuning using Unsloth Studio, supporting a context window of up to 262,144 tokens.

Finetunes
2 models