CAD Generation¶
3D Generation
Open Source¶
-
Arko-T Discovered by AI
2026
A 4-billion parameter text-to-design foundation model that maps natural-language intent directly into executable, parametric CAD programs by aligning every stage of the pipeline to a formal notion of design state.
cadfoundation-modelparametrictext-to-cad -
ASSEMCAD Discovered by AI
2026
A framework designed for generating production-ready mechanical CAD assemblies from natural language descriptions by coordinating multiple components, functional interfaces, and physical constraints.
assemblyb-repcadtext-to-cad -
CADAM Discovered by AI
2026
An open-source, browser-based text-to-CAD web application that leverages AI to transform natural language and images into editable 3D CAD models.
aicadtext-to-cadweb-app -
CADSmith Discovered by AI
2026
A multi-agent pipeline that generates manufacturing-ready CAD models from natural language by combining LLM code generation with programmatic geometric validation. It utilizes nested correction loops to resolve execution errors and verify dimensional accuracy using the OpenCASCADE kernel.
geometric-validationmulti-agentopencascadetext-to-cad -
Clarify Before You Draw Discovered by AI
2026
A proactive agentic framework for robust text-to-CAD generation that asks targeted clarifying questions to resolve under-specified or inconsistent user prompts before generating CadQuery scripts.
agentcadcadquerytext-to-cad -
DualBrep Discovered by AI
2026
A generative framework for B-Rep modeling that introduces a dual-field continuous representation to handle the heterogeneous structure of continuous parametric geometry and discrete topological graphs.
b-repcadgeometrytopology -
FutureCAD Discovered by AI
2026
A framework that integrates large language models with a boundary representation grounding transformer (BRepGround) to generate high-fidelity parametric CAD models from text descriptions.
brepcadparametrictext-to-cad -
Graph-CAD Discovered by AI
2026
A text-to-CAD generation framework that introduces a hierarchical, geometry-aware graph as an intermediate representation to explicitly model assembly structures and geometric constraints. This approach prevents cascading errors in long-horizon CAD code generation by structuring the design process before rendering.
assemblygeometry-awaregraph-representationtext-to-cad -
HierCAD: Hierarchical Text-to-CAD Design via Structure Alignment and Parameter Grounding Discovered by AI
2026
A hierarchical text-to-CAD framework that reformulates CAD generation as progressive reasoning by decomposing CAD construction trees into object-level procedural reasoning and part-level topology reasoning trajectories.
cadhierarchicalllmtext-to-cad -
HiFi-BRep Discovered by AI
2026
A robust B-Rep generation framework featuring a topology-aware encoder that eliminates padding noise and a single-stage parallel decoder that jointly predicts geometry and topology with differentiable manifold constraints.
b-repcadcvprlatent-representation -
Mamba-CAD Discovered by AI
2026
A self-supervised generative modeling framework for complex, industrial CAD models based on State Space Models (Mamba) to handle longer parametric CAD sequences.
cadmambaparametricstate-space-model -
ParaCAD: Autoregressive B-Rep Shape Generation with Parametric Surfaces Discovered by AI
2026
An autoregressive framework for point-cloud-conditioned B-Rep generation that directly operates on native parametric surfaces, recovering valid CAD models through global surface intersection and optimization.
autoregressivebrepcadpoint-cloud -
Pointer-CAD: Unifying B-Rep and Command Sequences via Pointer-based Edges & Faces Selection Discovered by AI
2026
An LLM-based text-to-CAD framework that treats CAD generation as a multi-step dialogue between a textual prompt and the evolving B-Rep (Boundary Representation), predicting operations while directly pointing to geometric entities.
brepcadparametrictext-to-cad -
STEP-LLM Discovered by AI
2026
A framework that fine-tunes compact LLMs to directly generate valid ISO 10303-21 STEP files from natural language descriptions, bypassing kernel-dependent script formats.
cadllmsteptext-to-cad -
SynapsCAD Discovered by AI
2026
An open-source AI-powered 3D CAD IDE that allows users to edit OpenSCAD code, visualize models in 3D, and reshape designs using natural language.
ai-assistedcadideopenscad -
Zero-to-CAD Discovered by AI
2026
An agentic framework developed by Autodesk Research that uses an LLM in a feedback-driven CAD environment to synthesize approximately one million executable CAD construction sequences with interpretable parameters.
agenticcadcadquerysynthetic-data -
A Solver-Aided Hierarchical Language for LLM-Driven CAD Design Discovered by AI
2025
Introduces AIDL (AI Design Language), a solver-aided, hierarchical domain-specific language that offloads spatial reasoning requirements to a geometric constraint solver, enabling large language models to generate complex parametric CAD designs.
caddslllmparametric -
AutoBrep: Autoregressive B-Rep Generation with Unified Topology and Geometry Discovered by AI
2025
An autoregressive Transformer model that generates high-quality Boundary Representation (B-Rep) CAD models by encoding both geometric and topological characteristics into a unified tokenization scheme.
autoregressiveb-repcadtransformer -
B-repLer: Semantic B-rep Latent Editor using Large Language Models Discovered by AI
2025
A text-driven editing framework for Boundary Representation (B-rep) CAD models that maps natural language instructions directly onto edits in a learned latent space.
brepcadeditinglatent-space -
BrepDiff: Single-stage B-rep Diffusion Model
2025
A single-stage diffusion model that directly generates boundary-representation (B-rep) CAD models, avoiding multi-stage pipelines for more robust parametric solid synthesis.
brepcaddiffusion -
BrepGPT: Autoregressive B-rep Generation with Voronoi Half-Patch Discovered by AI
2025
A single-stage autoregressive framework for B-rep CAD generation that introduces the Voronoi Half-Patch (VHP) representation to decompose irregular B-rep graphs into uniform tokens.
autoregressiveb-repcadvoronoi -
CAD-Coder Discovered by AI
2025
An open-source vision-language model explicitly fine-tuned to generate editable CAD code (CadQuery Python) directly from visual inputs, enabling precise 3D model reconstruction.
cadcadqueryimage-to-cadvlm -
CAD-Coder: Text-to-CAD Generation with Chain-of-Thought and Geometric Reward Discovered by AI
2025
A text-to-CAD generation framework that reformulates the task as producing CadQuery Python scripts, utilizing supervised fine-tuning and reinforcement learning with geometric rewards.
cadcadquerygrpotext-to-cad -
CAD-GPT: CAD Construction Sequences with Multimodal LLMs
2025
Synthesizes CAD construction sequences with a spatial-reasoning-enhanced multimodal LLM, generating editable modeling steps from text or single-image input.
cadllmmultimodalspatial-reasoning -
CAD-Recode: Reverse Engineering CAD Code from Point Clouds
2025
Reverse-engineers CAD models from point clouds by generating Python CAD code, recovering editable parametric construction programs from raw 3D scans.
cadcode-generationpoint-cloudreverse-engineering -
CAD2Program: 2D CAD Drawings to 3D Parametric Models
2025
A vision-language approach that reconstructs 3D parametric models from 2D CAD drawings, predicting editable shape programs from orthographic engineering views.
cadreconstructionvision-language -
CADDreamer Discovered by AI
2025
A novel approach for generating boundary representations (B-Rep) of CAD objects from a single-view image using a primitive-aware multi-view diffusion model.
b-repcaddiffusionsingle-view -
CADFusion Discovered by AI
2025
A text-to-CAD generation framework that leverages visual feedback to enhance the performance of large language models (LLMs) in generating CAD models from textual descriptions, alternating between sequential learning and visual feedback stages.
cadllmtext-to-cadvisual-feedback -
CADKnitter NEW Discovered by AI
2025
CADKnitter generates multi-component CAD assemblies from text and geometric guidance by explicitly modeling inter-part relationships and structural constraints.
cadcompositionalparametrictext-to-cad -
cadrille Discovered by AI
2025
A multi-modal CAD reconstruction framework that processes point clouds, multi-view images, and text inputs within a unified vision-language model, optimized using online reinforcement learning.
multimodalreconstructionreinforcement-learningvlm -
Drawing2CAD Discovered by AI
2025
A sequence-to-sequence learning framework designed to generate parametric 3D CAD models directly from 2D vector drawings (SVGs). It bridges the gap between traditional 2D engineering drawings and 3D CAD reconstruction by learning to predict CAD command sequences from vector graphics.
drawing-to-cadparametricsequence-to-sequencevector-graphics -
DTGBrepGen: A Novel B-rep Generative Model through Decoupling Topology and Geometry Discovered by AI
2025
A topology-geometry decoupled framework for CAD Boundary Representation (B-rep) generation that sequentially generates topologically valid and geometrically high-quality 3D models.
brepcadgeometrytopology -
FlexCAD: Controllable CAD Generation with Fine-tuned LLMs
2025
A controllable CAD generation method that fine-tunes large language models to produce parametric construction sequences, supporting versatile, hierarchy-aware editing and control.
cadcontrollablellm -
HoLa: B-Rep Generation using a Holistic Latent Representation
2025
Generates boundary-representation (B-rep) CAD solids from a holistic latent representation that jointly encodes geometry and topology, improving the validity of synthesized parametric models.
brepcaddiffusion -
Multi-Agent CAD (MAC) Discovered by AI
2025
A decoupled multi-agent framework designed for efficient text-to-CAD generation using constrained test-time compute. It coordinates multiple specialized AI agents to generate highly accurate, printable 3D models from concise natural language prompts while significantly reducing token usage and inference costs.
3d-printingbuild123dmulti-agenttext-to-cad -
NURBGen Discovered by AI
2025
A text-to-CAD framework that generates high-fidelity 3D CAD models directly from text using Non-Uniform Rational B-Splines (NURBS) by fine-tuning an LLM to output NURBS surface parameters.
cadllmnurbstext-to-cad -
Seek-CAD Discovered by AI
2025
A training-free generative framework for 3D parametric CAD modeling that utilizes locally deployed DeepSeek-R1 with a retrieval-augmented generation (RAG) strategy and stepwise visual feedback.
caddeepseekragtraining-free -
2025
Generates CadQuery Python code from text prompts using large language models, so natural-language descriptions compile into editable parametric CAD models.
cadcode-generationllmtext-to-3d -
BlenderLLM Discovered by AI
2024
A framework for training large language models specifically for CAD tasks using a self-improvement methodology. It generates executable Python scripts to construct and render 3D CAD geometry within Blender.
blendercadcode-generationllm -
2024
Directly generates boundary representation (B-rep) CAD solids using a diffusion model over a structured tree of geometric and topological entities.
brepcaddiffusion -
CAD-MLLM Discovered by AI
2024
A unified Computer-Aided Design generation system based on Multimodal Large Language Models (MLLMs) that can generate parametric CAD models conditioned on text, images, point clouds, or their combinations.
cadmllmmultimodalparametric -
CadVLM: Bridging Language and Vision for Parametric CAD Sketches
2024
A vision-language model that generates and autocompletes parametric CAD sketches, unifying visual and linguistic cues for sketch construction.
cadsketchvision-language -
Don't Mesh with Me: CSG Generation via Code-Gen LLM
2024
Fine-tunes a code-generation LLM to output Constructive Solid Geometry (CSG) programs instead of meshes, producing compact, editable primitive-based 3D shapes.
cadcode-generationcsgllm -
GenCAD Discovered by AI
2024
An image-conditioned CAD generation model that leverages a transformer-based contrastive representation and diffusion priors to generate both the 3D solid and the entire parameterized CAD command history.
caddiffusionimage-to-cadparametric -
Img2CAD NEW Discovered by AI
2024
Img2CAD factorizes single-view image-to-CAD reconstruction into 3D structure generation and VLM-guided CAD sequence recovery, producing editable parametric CAD programs.
cadimage-to-cadreverse-engineeringvlm -
text-to-cad Discovered by AI
2024
A library of agent skills for generating, inspecting, sourcing, slicing, and handing off CAD and robot-description artifacts from local project files, supporting STEP, STL, 3MF, and GLB exports.
agentscadstepstl -
2024
Generates parametric CAD construction sequences from natural language text descriptions of varying levels of detail.
cadtext-to-3dtransformer -
Text2CAD Discovered by AI
2024
An end-to-end Transformer-based autoregressive framework for generating complete, parametric CAD design histories from natural language prompts across beginner-to-expert skill levels.
caddeepcadparametrictext-to-cad -
PlankAssembly: 3D Reconstruction from Orthographic Views
2023
Reconstructs 3D shapes from three orthographic views by learning shape programs, robustly recovering editable structure even from noisy or inconsistent engineering drawings.
cadreconstructionshape-program -
SolidGen: Autoregressive B-rep Synthesis
2023
An autoregressive model that directly synthesizes boundary-representation (B-rep) solids, generating vertices, edges, and faces without relying on construction-sequence supervision.
autoregressivebrepcad -
ComplexGen: CAD Reconstruction by B-Rep Chain Complex Generation
2022
Reconstructs CAD models as a B-rep chain complex, jointly generating geometric primitives and their topological connectivity from input point clouds.
brepcadreconstruction -
Free2CAD: Parsing Freehand Drawings into CAD Commands
2022
Parses freehand design drawings into sequential CAD modeling commands, turning rough sketches into editable parametric construction operations.
cadfreehandsketch -
Point2Cyl: Reverse Engineering to Extrusion Cylinders
2022
Reverse-engineers 3D objects into editable extrusion cylinders, decomposing point clouds into sketch-extrude operations that recover parametric CAD structure.
cadreverse-engineeringsketch-extrude -
Reconstructing Editable Prismatic CAD from Voxels
2022
Recovers editable prismatic CAD models from rounded voxel inputs, converting approximate volumetric shapes into clean parametric solids with sketch-extrude structure.
cadreconstructionsketch-extrude -
2022
An autoregressive generative model for CAD sketch-and-extrude construction sequences using disentangled codebooks for diverse shape synthesis.
autoregressivecadsketch-extrude -
Vitruvion: A Generative Model of Parametric CAD Sketches
2022
A generative model of parametric CAD sketches that autoregressively synthesizes primitives and constraints, enabling sketch generation, completion, and conditional inference.
autoregressivecadsketch -
Computer-Aided Design as Language
2021
Casts parametric CAD construction as a language-modeling problem, generating CAD sketch-and-constraint sequences with a transformer over a domain-specific grammar.
cadlanguage-modelsketch -
2021
A generative model of CAD command sequences that produces editable parametric CAD models rather than raw geometry.
autoregressivecadsequence -
Engineering Sketch Generation for CAD
2021
Generates engineering sketches—the 2D constrained profiles underlying CAD models—using sequence models, a building block for downstream solid construction.
cadgenerationsketch -
SketchGen: Generating Constrained CAD Sketches
2021
An autoregressive transformer that generates constrained CAD sketches, producing primitives together with the geometric constraints that make them editable parametric models.
cadconstraintssketch -
Sketch2CAD: Sequential CAD Modeling by Sketching in Context
2020
Enables sequential CAD modeling by interpreting user sketches in context, mapping each sketched gesture to a parametric CAD operation for interactive construction.
cadinteractivesketch
Proprietary / Industry Tools¶
-
Adam CAD Copilot Discovered by AI
2026
An AI-powered CAD copilot designed for hardware and mechanical engineering teams that integrates directly into professional CAD software like Onshape and Autodesk Fusion. It supports prompt-driven part editing, selection-based feature creation, and automated feature tree optimization.
cadcopilotfusion360onshape -
Ragnar CAD Discovered by AI
2025
An AI-powered CAD platform that converts text descriptions and hand-drawn sketches into precise, editable 3D CAD models (B-Rep geometry) with STEP and STL export.
brepcadsketch-to-cadtext-to-cad -
SGS-1 (Spectral Labs) Discovered by AI
2025
A spatial-intelligence foundation model that generates fully manufacturable, parametric 3D CAD geometry in B-Rep (STEP) format from images, 3D meshes, or text descriptions.
b-repcadparametricstep -
Zoo Design Studio Discovered by AI
2025
A professional CAD platform and infrastructure for hardware design that features built-in Text-to-CAD capabilities and Zookeeper, a conversational design agent for generating editable B-Rep CAD models from natural language.
brepcadconversational-aitext-to-cad -
Zoo Text-to-CAD Discovered by AI
2024
A conversational CAD agent and API (Zookeeper) that generates production-ready, editable B-Rep CAD files and meshes from natural language prompts, bypassing traditional mesh-only 3D generation.
apibrepcadtext-to-cad -
Zoo Zookeeper Discovered by AI
2024
A conversational CAD agent built into Zoo Design Studio that generates precise, fully editable B-rep CAD models from natural language prompts.
b-repcadproprietarytext-to-cad -
Leo AI Discovered by AI
2023
A patented Large Mechanical Model (LMM) and mechanical engineering copilot that generates full CAD assemblies, parametric models, and engineering specifications from text prompts or sketches.
assemblycadcopilotlmm -
Zoo Discovered by AI
2023
A developer-focused generative CAD platform (formerly KittyCAD) that converts natural language prompts into precise, editable B-Rep CAD models in formats like STEP, STL, and OBJ.
apib-repcadtext-to-cad
