Large Language Models for Computer-Aided Design: A Survey

TL;DR

This survey addresses the gap in understanding how Large Language Models (LLMs) can augment Computer-Aided Design (CAD) workflows. It surveys CAD's industrial significance, then distills foundational LLM concepts, including Transformer-based architectures, training pipelines, alignment, and prompting, before detailing state-of-the-art closed-source and publicly available models relevant to CAD. A six-part taxonomy (data generation, CAD code generation, parametric CAD generation, image generation, model evaluation, and text generation) organizes CAD-oriented LLM research, highlighting methods that rely on intermediate representations like code and JSON rather than direct 3D outputs, and emphasizing multimodal inputs. Key findings indicate strong use of GPT-family models and a trend toward intermediate-data generation plus multimodal integration, with future directions spanning interior design, data-format generation, AEC compliance, and fashion/textile CAD, all with significant potential to automate and enhance CAD workflows. Notably, discussions reference large-scale model characteristics such as up to $405B$ parameters and context lengths up to $128K$ tokens, illustrating the scale at which these systems operate and the opportunities for CAD-enabled efficiency gains.

Abstract

Large Language Models (LLMs) have seen rapid advancements in recent years, with models like ChatGPT and DeepSeek, showcasing their remarkable capabilities across diverse domains. While substantial research has been conducted on LLMs in various fields, a comprehensive review focusing on their integration with Computer-Aided Design (CAD) remains notably absent. CAD is the industry standard for 3D modeling and plays a vital role in the design and development of products across different industries. As the complexity of modern designs increases, the potential for LLMs to enhance and streamline CAD workflows presents an exciting frontier. This article presents the first systematic survey exploring the intersection of LLMs and CAD. We begin by outlining the industrial significance of CAD, highlighting the need for AI-driven innovation. Next, we provide a detailed overview of the foundation of LLMs. We also examine both closed-source LLMs as well as publicly available models. The core of this review focuses on the various applications of LLMs in CAD, providing a taxonomy of six key areas where these models are making considerable impact. Finally, we propose several promising future directions for further advancements, which offer vast opportunities for innovation and are poised to shape the future of CAD technology. Github: https://github.com/lichengzhanguom/LLMs-CAD-Survey-Taxonomy

Figures (4)

• Figure 1: The taxonomy of this review.
• Figure 2: Typical CAD code generation pipeline. A prompt and, optionally, an image are first input to a VLM or LLM. The VLM/LLM then generates the corresponding code, which is executed. If the code is not executable, refinement is performed until it is. Subsequently, the generated objects' computing similarity to the ground truth is evaluated. If the similarity is below a threshold, the code refinement process is repeated until the desired results are achieved.
• Figure 3: A standard parametric CAD generation pipeline. A prompt, along with an optional image (converted into features by a frozen image encoder), is fed into a trainable VLM or LLM to generate parametric data. This data is then parsed to produce 3D CAD models. LLMs can also be employed to compute the similarity between the ground truth and the generated object.
• Figure 4: Bar chart of LLM usage frequency in CAD-related research (sorted descending).