emhltbkars/xxx

VISIONConcurrent Unit Cost:1Model Size:9BQuant:FP8Context Size:32kTool Calling:SupportedPublished:Aug 11, 2026License:mitArchitecture:Transformer Open Weights Featherless Exclusive Cold

Ornith-1.0-9B is a 9 billion parameter model developed by DeepReinforce Team, part of 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 excels in coding benchmarks such as Terminal-Bench 2.1, SWE-Bench, NL2Repo, and OpenClaw, offering state-of-the-art performance among open-source models of comparable size. It is optimized for efficient single-GPU deployment and features reasoning and tool-calling capabilities.

Loading preview...

Ornith-1.0-9B: Agentic Coding Model

Ornith-1.0-9B, developed by the DeepReinforce Team, is a 9 billion parameter model from the Ornith-1.0 family, focused on agentic coding. It stands out for its self-improving training framework which uses reinforcement learning to generate and optimize both solution rollouts and the underlying scaffolds, leading to higher-quality solutions.

Key Capabilities & Features

  • State-of-the-Art Coding Performance: Achieves leading results on benchmarks like Terminal-Bench 2.1 (43.1 on Terminus-2, 40.6 on Claude Code), SWE-Bench Verified (69.4), SWE-Bench Pro (42.9), NL2Repo (27.2), and Claw-eval Avg (63.1), often outperforming larger models like Qwen3.5-35B and Gemma4-31B in its size class.
  • Reasoning Model: By default, the model generates a <think> ... </think> block for chain-of-thought reasoning before providing the final answer, which can be parsed separately.
  • Tool-Calling: Emits well-formed function calls that are parsed into standard OpenAI-style tool_calls, making it compatible with various agent frameworks.
  • Efficient Deployment: Designed for single-GPU deployment, making it accessible for local inference and fine-tuning.
  • Open-Source License: MIT licensed, ensuring global accessibility and freedom from regional limitations.

Good For

  • Developing coding agents: Its strong performance on agentic coding benchmarks makes it suitable for building automated coding assistants.
  • Automating software development tasks: Excels at understanding large codebases and automating tedious work.
  • Integration with agent frameworks: Compatible with standard frameworks like Hermes Agent, OpenClaw, and OpenHands due to its OpenAI-compatible API and tool-calling capabilities.
  • Local development and experimentation: Its efficient design allows for comfortable deployment on a single 80GB GPU.

Finetunes
1 models