minghinmatthewlam/openbench

minghinmatthewlam/openbench provides a ready‑made benchmark harness that lets you compare coding agents (Codex, PI, OpenCode, Cursor, Devin) and open models on