Ornith-1.0: Self-Scaffolding LLMs for Agentic Coding

Benchmark Model Rank Results
code-generation-on-swe-bench-proOrnith-1.0-397B#1Resolved: 62.2
code-generation-on-terminal-bench-2-1Claude Code + Ornith-1.0-397B#2Accuracy: 78.2
code-generation-on-terminal-bench-2-1Terminus-2 + Ornith-1.0-397B#3Accuracy: 77.5
code-generation-on-terminal-bench-2-1Terminus-2 + Ornith-1.0-35B#6Accuracy: 64.2
code-generation-on-terminal-bench-2-1Claude Code + Ornith-1.0-35B#9Accuracy: 62.8
code-generation-on-terminal-bench-2-1Terminus-2 + Ornith-1.0-9B#13Accuracy: 43.1
code-generation-on-terminal-bench-2-1Claude Code + Ornith-1.0-9B#14Accuracy: 40.6