GPT 6 Astra shows promise and limits in CAD tasks
Assessing the CAD Capabilities of General-Purpose AI Models: A Case Study of GPT-6 Astra
Graphics
Summary
Generating and editing technical 3D design files is hard for AI because they must be precise and functional, not just look right. The authors studied how well GPT-6 Astra can do basic computer-aided design (CAD) work, such as understanding, creating, changing, and checking models. They found it can support many steps in design, even for complex shapes, but still makes some critical errors that affect usability. Fixing problems with repeated attempts helps but is slow and not always successful.
What this means in practice
- •For mechanical engineers: Assist in creating and checking CAD parts with partial automation while identifying common design errors specific to geometry and constraints.
- •For product design teams: Use AI-guided iterative model refinement to speed early-stage design exploration despite remaining functional limitations in generated CAD data.
Authors
Qiang Zou, Jiajun Shen, Zhijie Yang, Yuhao Wang, Shitianyi Leng, Dongshuo Yin
Abstract
Recent advances in general-purpose AI models and agentic systems have enabled them to perform some CAD tasks. A systematic assessment is now needed to establish what these systems can accomplish and how well their outputs meet engineering requirements. This paper presents an empirical study of GPT-6 Astra, a SOTA general-purpose AI model, across four basic CAD tasks: model understanding, generation, editing, and evaluation. We design test cases spanning different levels of complexity and engineering requirements. The results show that GPT-6 Astra can support multiple stages of CAD workflows, with successful outcomes in cases involving nontrivial geometric complexity. This suggests that general-purpose AI is beginning to offer practical value in almost all steps in CAD workflow. However, gaps remain between generating visually plausible geometry and producing valid, functional, and editable CAD models. Recurring errors in freeform geometry, parameter dependencies, and mating relationships between parts limit engineering usefulness. Agentic iterative refinement can improve results but does not consistently resolve critical errors; long completion times further constrain practical use.