Ornith 1.0 35B

Ornith-1.0-35B is a text-generation model that serves as the lightweight single-GPU member of the Ornith family, a self-improving family of open-source models for agentic coding post-trained on Gemma 4 and Qwen 3.5 architectures.

Training Framework

The model employs a self-improving reinforcement learning framework that jointly optimizes both the scaffolding (search trajectories) and the resulting solution rollouts. By learning to generate better scaffolds alongside solutions, Ornith-1.0-35B discovers higher-quality search trajectories and produces improved coding solutions. The model is released under the MIT license.

Key Capabilities

Ornith-1.0-35B is a reasoning model: by default, the assistant response opens with a <think> … </think> block before delivering the final answer. It achieves strong results across multiple agentic coding benchmarks.

Benchmark Results

	Ornith-1.0-35B	Qwen3.5-35B	Qwen3.6-35B	Gemma4-31B	Qwen3.5-397B
Agentic Coding
Terminal-Bench 2.1 (Terminus-2)	64.2	41.4	52.5	42.1	53.5
Terminal-Bench 2.1 (Claude Code)	62.8	38.9	49.2	-	48.6
SWE-bench Verified	75.6	70	73.4	52	76.4
SWE-bench Pro	50.4	44.6	49.5	35.7	51.6
SWE-bench Multilingual	69.3	60.3	67.2	51.7	69.3
NL2Repo	34.6	20.5	29.4	15.5	36.8
Claw-eval Avg	69.8	65.4	68.7	48.5	70.7
SWE Atlas - QnA	37.1	13.2	15.5	-	20.4
SWE Atlas - RF	29.7	10.2	11.4	-	18.4
SWE Atlas - TW	27.8	9.8	13.3	-	18.5

Bold numbers indicate the highest score in each row among the compared models. Evaluation details: Terminal-Bench 2.1 uses Harbor/Terminus-2 and Claude Code frameworks with temperature 1.0; SWE-Bench uses OpenHands harness; SWE Atlas uses mini SWE agent harness; NL2Repo uses 400K context; ClawEval uses temperature 0.6 with 256K context. All results averaged over 5 runs where noted.