I tested 10 model/harness combinations on the same Three.js task

The author evaluated ten different model-and-harness pairings by prompting each to generate a self-contained HTML page that builds a sci-fi hangar in Three.js with hovering drones, animated warning lights, emissive runway strips, volumetric fog, a drone-formation toggle, and a cinematic camera path. Results were recorded in “/goal” mode, noting total output tokens (including reasoning where reported), GLM 5.3 Flash Max labeling, DSH adapter token handling, and tool-error counts, with unavailable metrics marked by dashes.