Antigravity 2.0 has topped the OpenSCAD Architectural 3D LLM Benchmark by successfully generating a detailed parametric CAD model of the Pantheon, according to modelrift.com. The benchmark, conducted recently, tested several AI coding tools on their ability to convert architectural reference images into OpenSCAD code for 3D modeling.
The benchmark involved giving AI systems the same task: build the Pantheon using OpenSCAD, a script-based CAD modeling language. The models were evaluated on their capacity to render complex spatial geometry, including the rotunda, dome, portico, columns, pediment, and other front details. Antigravity 2.0 outperformed competitors such as Codex 5.5 High, Claude Sonnet, Claude Opus, Cursor Composer, and Google Antigravity by producing the most accurate and detailed parametric code, as verified by rendered previews.
This benchmark is significant because it highlights advancements in AI's ability to handle architectural and spatial design tasks, a challenging area for language models. Accurate parametric CAD code generation can accelerate architectural design workflows and improve automation in 3D modeling. The competition among leading AI coding tools reflects growing interest in integrating AI with CAD software, which could reshape how architects and engineers approach design.
Looking ahead, the benchmark will likely influence further development of AI models specialized in 3D spatial reasoning and CAD code generation. ModelRift plans to continue using this benchmark to track improvements and ship enhanced 3D models on its platform. Observers should watch for updates on how these AI tools evolve to support increasingly complex architectural and engineering projects.