In 1975, Martin Newell, a computer graphics researcher at the University of Utah, needed a simple 3D object to test his rendering algorithms. His wife suggested he model a teapot. She was right. Newell drew a Melitta teapot on graph paper, entered the control points into his computer as a set of Bezier patches, and created a 3D model that became the most famous object in computer graphics history. The Utah teapot has appeared in hundreds of research papers, software Easter eggs, and animated films. It is the "Hello World" of 3D modeling. Newell's teapot was not art. It was a test case. But it established that a physical object could be represented as a set of mathematical surfaces in a computer, and that those surfaces could be rendered as a recognizable image. Every 3D model made since, from Pixar characters to game environments to AI-generated meshes, descends from that teapot.
3D modeling is the process of creating three-dimensional objects in digital space. A 3D model is a mathematical representation of a physical or imaginary object, stored as a collection of points, surfaces, and textures that a computer can render as a two-dimensional image. The model exists in a virtual coordinate system with three axes (x, y, z), and it can be viewed from any angle, lit from any direction, and animated over time. 3D modeling is used in film, video games, architecture, product design, scientific visualization, and fine art.
This entry covers the history of 3D modeling from the 1970s to AI-generated meshes in 2026, the artists and researchers who defined the field, and how new tools are making 3D modeling accessible to people who have never touched a 3D application.
What Is 3D Modeling and How Does It Work?
There are several approaches to 3D modeling, each with its own tools and use cases. Polygonal modeling represents objects as meshes of triangles or quadrilaterals. Each polygon has vertices (points in 3D space), edges (lines connecting vertices), and faces (the surfaces bounded by edges). A simple cube has 8 vertices, 12 edges, and 6 faces. A complex character model may have millions of polygons. Polygonal modeling is the most common approach in game development and real-time applications because GPUs are optimized to render polygon meshes efficiently.
NURBS modeling (Non-Uniform Rational B-Splines) represents surfaces as mathematical curves defined by control points. NURBS surfaces are smooth and continuous, unlike polygon meshes which are faceted at low resolutions. NURBS are used in automotive design, industrial design, and animation where smooth, manufacturable surfaces are required. The Utah teapot was modeled as a set of Bezier patches, which are a type of NURBS surface.
Digital sculpting, popularized by ZBrush (released in 1999 by Pixologic), lets artists push, pull, and smooth a virtual clay surface using a pressure-sensitive tablet. The artist starts with a simple shape and builds up detail by sculpting, much like working with physical clay. ZBrush models can have billions of polygons, far more than real-time engines can handle, so the high-resolution sculpture is typically "baked down" to a lower-resolution mesh with a normal map that preserves the appearance of fine detail.
Procedural modeling generates geometry algorithmically rather than by hand. The artist defines rules and parameters, and the software generates the model. This approach is used for environments, architecture, and objects that follow patterns: forests, cities, roads, buildings. Procedural modeling connects 3D modeling to generative art and to procedural generation.
Origins and History
The Pioneering Era (1960s-1970s)
3D modeling grew out of computer-aided design (CAD) research in the 1960s. Ivan Sutherland's Sketchpad (1963), which also founded vector art, was a 2D drafting system, but its constraint-based approach to geometric representation directly influenced 3D CAD systems. The first 3D solid modeling systems appeared in the early 1970s. PADL-1, developed at the University of Rochester by Herbert Voelcker and Ari Requicha, used constructive solid geometry (CSG) to represent objects as combinations of primitive shapes (spheres, cylinders, cubes) joined by Boolean operations (union, intersection, difference).
Martin Newell's Utah teapot (1975) was created at the University of Utah, which was one of the leading centers of computer graphics research in the 1970s. The Utah graphics lab, directed by David Evans and Ivan Sutherland (who joined Utah after MIT), produced a generation of researchers who founded or influenced nearly every major computer graphics company. Ed Catmull, who would later co-found Pixar, completed his PhD at Utah in 1974. His thesis introduced the Z-buffer, texture mapping, and the concept of rendering curved surfaces by subdividing them into polygons. These three ideas are the foundation of real-time 3D rendering.
The CG Film Era (1980s-2000s)
The 1980s saw 3D modeling move from research labs to film production. Industrial Light and Magic (ILM) used 3D CGI in films including Star Trek II: The Wrath of Khan (1982, the Genesis Effect sequence) and The Abyss (1989, the water tentacle). Pixar, spun off from Lucasfilm's computer division in 1986, developed its own modeling and rendering software (Marionette and RenderMan) for animated shorts and eventually feature films. Toy Story (1995), directed by John Lasseter, was the first feature-length film made entirely with 3D computer graphics. Every character, set, and prop was modeled in 3D using polygonal modeling and NURBS surfaces.
The 1990s also saw 3D modeling enter game development. id Software's Quake (1996) was one of the first games to use fully 3D models for characters and environments, rendered in real time. The models were low-polygon by necessity: Quake's characters had a few hundred polygons each, compared to the millions used in film. This gap between film-quality and game-quality 3D modeling would persist for two decades, narrowing as GPU power increased.
The Democratization Era (2000s-2020s)
Blender, first released in 1998 and open-sourced in 2002, brought professional-grade 3D modeling to anyone with a computer. Blender's development was funded by community donations and later by the Blender Foundation, a nonprofit established in 2002. By 2026, Blender is used by independent artists, small studios, and even major productions, and its feature set rivals commercial tools like Maya and 3ds Max.
ZBrush, released in 1999, revolutionized digital sculpting. It allowed artists to create models with organic, sculpted detail that would have been impractical to achieve with traditional polygonal modeling. ZBrush became the standard tool for character modeling in film and games, and its workflow influenced competing tools like Mudbox (Autodesk, 2006) and Blender's Sculpt mode.
Key Artists and Pioneers
Martin Newell (b. 1947)
Newell created the Utah teapot in 1975 at the University of Utah. The teapot was modeled as 32 Bezier patches and has been used as a standard test object in computer graphics for 50 years. Newell later worked at CADlab and Newell's Rubbermaid, but the teapot remains his most recognized contribution. It is in the collection of the Computer History Museum in Mountain View, California.
Ed Catmull (b. 1945)
Catmull's PhD thesis at the University of Utah (1974) introduced the Z-buffer (for hidden surface removal), texture mapping (for applying 2D images to 3D surfaces), and a method for rendering curved surfaces by subdividing them into polygons. These three innovations are used in every 3D rendering system today. Catmull co-founded Pixar in 1986 and served as its president until 2019. He received the Turing Award in 2019 for his contributions to 3D computer graphics.
Robert Lazzarini (b. 1965)
Lazzarini is a contemporary artist who uses 3D modeling and digital fabrication to create sculptures that distort physical space. His payphone (2006) is a 3D-modeled payphone that has been mathematically distorted and then fabricated in physical materials at the same scale as a real payphone. The sculpture looks like a 3D model rendered with a wrong shader, but it is a physical object. Lazzarini's work demonstrates that 3D modeling is not limited to screen-based art. It can generate forms that are then materialized as sculpture.
Kevin Mack (b. 1959)
Mack won the Academy Award for Best Visual Effects in 2000 for his work on What Dreams May Come (1998), which used 3D modeling to create a painted world that the characters move through. Mack's more recent work uses 3D modeling and digital fabrication to create abstract sculptures that he describes as "computational art." His sculptures are modeled in 3D software and then 3D-printed in materials including bronze and resin. Mack's work bridges the gap between 3D modeling as a production tool and 3D modeling as a fine art medium.
AI-Generated 3D Models (2026)
In 2026, AI 3D generation has moved from research curiosity to practical tool. Meta released SAM 3D at CVPR 2026, a foundation model that predicts 3D object geometry, texture, and pose from a single image. SAM 3D was trained on nearly 1 million annotated images with over 3.14 million untextured meshes and 100,000 textured meshes, an unprecedented scale for 3D data. The model achieves at least a 5:1 win rate in human preference tests over prior work on real-world objects. Read the paper at CVPR 2026.
Also at CVPR 2026, researchers presented MeshFlow, a method for generating artist-quality 3D meshes in approximately one second. MeshFlow uses a Variational Autoencoder to compress mesh data (vertices, normals, connectivity) into a continuous latent space, then generates meshes using a rectified flow transformer. It is 18 times faster than the fastest autoregressive mesh generator while achieving better accuracy. Read the paper at CVPR 2026.
Roblox announced in July 2026 a suite of AI-powered 3D creation tools, including Build, a mobile-first creation tab that turns text prompts into basic game environments, and Cube, a 3D foundation model that generates game-ready objects from prompts. Roblox's Procedural Models generate parametric 3D assets from text or image input, with adjustable parameters for reshaping without manual modeling. Read more at Roblox.
Adobe announced Substance 3D updates at SIGGRAPH 2026 in July, including Substance 3D Painter 12.1 with OpenPBR material support, a redesigned baking workspace, and Hard Surface Auto UV mode. Substance 3D Designer 16 added new Signed Distance Field nodes for procedural 3D modeling directly in the node graph. Read more at Adobe.
3D modeling is the first step in the 3D graphics pipeline. Once a model exists, it must be rendered to produce a 2D image, and its surfaces must be described by shaders that determine how light interacts with them. 3D modeling connects to procedural generation, where models are created algorithmically rather than by hand. It relates to digital art as one of its major disciplines, and to generative art when the modeling process is driven by code or AI. For more on digital art, read our post on the history of digital art from pixel to prompt, or explore our guide to digital art.
See 3D Modeling in Practice
3D modeling is visible in every CGI film, every 3D video game, and every architectural visualization. To see 3D modeling as fine art, look at Robert Lazzarini's sculptures, exhibited at galleries including the Museum of Modern Art in New York. Kevin Mack's computational sculptures are exhibited at galleries specializing in digital and new media art. The Utah teapot is in the Computer History Museum in Mountain View, California.
For more on the 3D pipeline, read our entries on rendering and shaders, or explore our post on the history of digital art.
