Arko-T: A Foundation Model for Text-to-Structured 3D Generation

2026-06-29

AI Summary

Existing text-to-3D generation methods produce non-editable meshes and lack the capability to generate parametric CAD designs. This work proposes the first end-to-end foundational model (4B parameters) that directly maps natural language instructions to executable, editable CAD modeling programs. By leveraging large-scale language modeling, code normalization, execution-based feedback supervision, and design-state alignment mechanisms, the model maintains consistency across geometric structure, parametric values, and modeling logic. It achieves state-of-the-art performance, ranking first on 8 out of 12 evaluation metrics and second on 3 others, matching or surpassing leading general-purpose large models while requiring only approximately one-tenth of their inference cost per evaluation.

Abstract

Text-to-3D systems can now synthesize a mechanical part from a single sentence, yet the result is a shape to render, not a design to edit. We present Arko-T, a 4B-parameter text-to-design model that maps natural-language intent directly into executable, parametric CAD programs. Rather than optimizing for code executability alone, Arko-T aligns every stage of the pipeline to a formal notion of design state, so that data curation, code normalization, and execution-grounded supervision all work to preserve the features, parameters, and construction logic that make a CAD artifact editable. Benchmarked against seven frontier LLMs across 12 metrics, Arko-T attains the best score on 8 and the second-best on 3 more, at roughly one-tenth the per-benchmark cost. The results suggest that targeted design-level training at moderate scale can match frontier general-purpose models on structured CAD generation.

Problem

• Text-to-3D
• parametric CAD
• structured 3D generation
• editable design
• foundation model

Innovation

• text-to-CAD
• parametric modeling
• design state alignment
• executable CAD programs
• foundation model