Reported agentic-coding scores
According to the model card, the model posts publisher-reported scores such as 43.1 on Terminal-Bench 2.1 (Terminus-2) and 69.4 on SWE-bench Verified.
Open Source Model Profile · unsloth
Ornith-1.0-9B
Ornith-1.0-9B is Unsloth's lightweight Qwen3.5-family text-generation model for agentic coding. According to the model card, it targets efficient single-GPU deployment with tool-calling support and publisher-reported coding-benchmark scores.
- Publisher
- unsloth
- Task
- text-generation
- Model type
- qwen3_5
- License
- mit
- Library
- Unknown
- Publication status
- Accepted · not indexed
Model overview
Ornith-1.0-9B is published by Unsloth as a text-generation model for agentic coding. The captured configuration identifies Qwen3_5ForConditionalGeneration with model type qwen3_5. According to the model card, it is the most lightweight, single-GPU-oriented member of the Ornith-1.0 family, which is described as post-trained on Qwen 3.5 and Gemma 4 foundations with reinforcement-learned self-improvement.
Recorded capabilities
Documented tool calling
According to the model card, the model emits well-formed function calls parsed into the standard tool_calls field, supporting agentic coding workflows.
Single-GPU serving recipes
According to the model card, vLLM, SGLang, and Transformers recipes are documented, with the dense bf16 model described as fitting comfortably on a single 80GB GPU.
Self-improving training description
According to the model card, Ornith-1.0 uses reinforcement learning to jointly optimize solution rollouts and the scaffolds driving them.
Use cases in the source record
- Agentic coding experiments of the kind covered by the card's Terminal-Bench, SWE-bench, and NL2Repo results.
- Tool-calling assistant deployments served through an OpenAI-compatible vLLM or SGLang endpoint.
- Single-GPU local inference or fine-tuning setups using the documented Transformers and Unsloth recipes.
Limitations and unknowns
- Captured Safetensors metadata reports 1.47M parameters while the model card describes a dense ~9B model; this discrepancy needs review before parameter claims are presented as settled.
- No context-window value was extracted from this record.
- Benchmark figures are publisher-reported claims and were not independently verified.
- Provider state is historical snapshot data, not independently refreshed current availability.
Source and provenance
Source: unsloth/Ornith-1.0-9B
Captured: Unknown. Processed: 2026-09-07T19:35:00.193897+00:00.
Unsloth Dynamic 2.0 achieves superior accuracy & outperforms other leading quants. Ornith-1.0-9B Aloha! 🌺 Today, we are releasing Ornith-1.0, a self-improving family of open-source models for agentic coding. Highlights: State-of-the-Art Coding Agents : Available in 9B-Dense, 31B-Dense, 35B-MoE, and 397B-MoE (post-trained on top of Gemma 4 and Qwen 3.5), achieving 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. Self-Improving Training Framework : Ornith-1.0 employs RL to learn to generate not only solution rollouts, but also th…
F001F002F003F004F005F006F007F009F011F012F013F014F015F016F017F018F020F021F024F026