On this page
Three.js
Three.js
The dominant JavaScript library for 3D graphics in the browser, wrapping the low-level WebGL and (since r171) WebGPU APIs behind a developer-friendly scene-graph interface. The library handles cameras, lights, geometry, materials, animations, and model loading; retailers use it to build 360° product viewers, 3D configurators, and WebXR-based augmented reality experiences on product detail pages.
Market position (as-of 2026-07-14)
Three.js receives approximately 2.7 million weekly npm downloads, making it the clear leader in the web 3D space. [1]
Three frameworks dominate web-based 3D in ecommerce as of 2026 [2]:
| Framework | Primary ecommerce use case |
|---|---|
| model-viewer (Google) | Quick product embeds; minimal code; no custom logic |
| Three.js | Custom interactive scenes and 3D configurators |
| Babylon.js | Complex real-time 3D environments and game-style apps |
WebGPU transition (as-of 2026-06-17)
Three.js r171 (September 2025) introduced a production-ready WebGPU renderer (WebGPURenderer) with automatic fallback to WebGL 2 for older browsers; switching from the legacy WebGLRenderer requires a one-line import change. [3]
As of January 2026, WebGPU achieved Baseline status across Chrome, Edge, Firefox (Windows/macOS), and Safari 26 (iOS and macOS Tahoe) — shipping stable with no flags required. VR.org describes this as "the end of the 15-year WebGL era." (VR.org, 2026-05-06)
The 2025 Web Almanac put WebGPU adoption at 65% of new web apps shipping 3D content, up from approximately 8% two years prior. (as-of 2026-05-06; VR.org, 2026-05-06)
Babylon.js has been WebGPU-first since version 5.0 (2022), giving it an earlier head start on the WebGPU rendering path than Three.js. (VR.org, 2026-05-06)
Chrome 149–150 (June 2026) shipped WebGPU Immediates (push constants), allowing small amounts of frequently-changing per-draw-call data (e.g. 3D transformation matrices) to be passed directly to GPU shaders without CPU buffer overhead — a performance-relevant change for Three.js WebGPU renderer users. [4]
Ecommerce use cases
Cited use cases for Three.js / WebGL in ecommerce [5]:
- 360° product rotation viewers — furniture, appliances, cars; zoom, rotate, explore geometry
- 3D product configurators — real-time colour and material switching via
mesh.material.color; personalization of variants without page reload; feeds into KHR_materials_variants for glTF-native variant switching - Virtual try-on — eyewear, fashion via AR + WebGL overlay
- Mass-customization configurators — online printers and manufacturers enabling build-before-purchase flows
Named brand examples of 3D/AR deployment [6]:
- IKEA Place (in-space AR furniture placement)
- AutoTrader Canada (3D/AR vehicle browsing)
- Nike / Snapchat (AR try-on partnerships)
- Al Huzaifa (luxury furniture configurator)
- Sephora Virtual Artist (beauty AR)
London Dynamics (2025-02-28, updated 2025-07-10) — pre-2026 source included because no 2026 brand-level case study list was found in this pass.
Framework comparison
Coohom (April 2026) notes that "performance depends more on model optimization than the framework itself" — polygon count, texture resolution, and draw-call management matter more than library choice. [2]
GLB and GLTF are described as the most widely supported file formats for web-based 3D rendering across all three frameworks. (Coohom, 2026-04-25)
Babylon.js offers stronger built-in support for advanced glTF (GL Transmission Format) PBR material extensions (clearcoat, transmission, sheen) and often produces more accurate visual results for complex materials than Three.js out of the box (as-of 2026-04-25). (Coohom, 2026-04-25)
Production performance
Production Three.js for ecommerce — key considerations (AppScale Blog, 2026-07-14):
Top performance killers:
- Draw calls (not polygon count) are the primary bottleneck; remedies: geometry instancing, geometry merging, texture atlases
- Memory leaks from failing to call
.dispose()on geometries, materials, and textures when removing scene objects — leads to "memory-leak-until-crash" behaviour on long product-page sessions
Three-tier asset strategy:
- Draco compression for geometry
- KTX2 (GPU-native) textures
- Multiple Levels of Detail (LODs) for mobile — required because mobile fill-rate and thermal throttling mean desktop framerates rarely survive on real hardware
Progressive loading pattern:
- Serve a static poster image while the 3D scene loads (prevents blank-canvas experience)
- Attempt WebGPU renderer → fall back to WebGL 2 if unavailable
- (AppScale Blog, 2026-07-14)
React Three Fiber (R3F)
React Three Fiber is described by AppScale Blog as the mainstream abstraction layer for Three.js in React codebases as of 2026. Tutorial content shows Three.js also being wrapped for Vue/Nuxt (via TresJS) and Angular (via Angular Three). (AppScale Blog, 2026-07-14; YouTube sources, 2024–2025)
Dominant composable stack pattern in 2024–2025 tutorial content: Three.js + GSAP (scroll animation) + headless CMS (Prismic) + payment (Stripe) + meta-framework (Next.js or Nuxt 4). (Prismic YouTube tutorials, 2024-2025)
WebXR integration
As of 2026, Meta Quest Browser, Samsung Internet, and Safari on Apple Vision Pro expose WebXR through the same WebGPU pipeline, enabling write-once/hit-every-major-XR-runtime web experiences. (VR.org, 2026-05-06)
The Interop 2026 initiative (W3C browser vendor alignment) includes WebXR as a focus area, with new WebXR standards assumed to target a WebGPU rendering path. (VR.org, 2026-05-06)
Market size benchmarks
The following conversion/return benchmarks originate from 2020 primary data (Shopify via HBR and AR Insider) and are widely re-cited but have not been independently updated as of this research pass. Treat as directionally indicative only, not current benchmarks.
- London Dynamics (2025) cites "60% of shoppers more likely to buy with 3D/AR" (attributed to AP News / Globe Newswire, pre-2026)
- London Dynamics (2025) cites "94% higher conversion for AR vs. no AR" (Shopify data via HBR, 2020)
- London Dynamics (2025) cites "40% decrease in product returns from 3D visualization" (Shopify via AR Insider, 2020)
3D ecommerce market sizing (as-of 2026): $8.53B (2025) → $10.54B (2026) at 23.6% CAGR. [7]
VR.org (May 2026) frames WebGPU as delivering "2-10x raw performance gains over WebGL in heavy scenes"; Coohom (April 2026) states "performance depends more on model optimization than the framework itself." These are not irreconcilable — asset optimization matters for any renderer, but the underlying API ceiling also matters at scale and with very high draw-call counts.
Key terms
| Term | Meaning |
|---|---|
| WebGLRenderer | Three.js legacy renderer using the WebGL 1/2 API |
| WebGPURenderer | Three.js new renderer (r171+) using the WebGPU API; one-line migration from WebGLRenderer |
| Scene graph | Three.js's hierarchical structure: Scene → Mesh (Geometry + Material) + Camera + Light |
| Draw call | One GPU command to render a mesh; too many per frame is the top Three.js bottleneck |
| Draco | Google geometry-compression algorithm; reduces GLB file sizes 70–90% |
| KTX2 | GPU-native texture format; eliminates CPU-to-GPU texture decompression overhead |
| LOD | Level of Detail — serving lower-poly versions at distance or on constrained devices |
| R3F | React Three Fiber — React bindings for Three.js; mainstream React abstraction as of 2026 |
References
- AppScale Blog, 2026-07-14 — appscale.blog/en/blog/threejs-production-3d-web-2026-webgpu-realtime-standards
- Coohom, 2026-04-25 — www.coohom.com/article/three.js-vs-model-viewer-vs-babylon.js-for-website-3d-integration
- VR.org, 2026-05-06 — vr.org/articles/webgpu-baseline-2026-three-js-webxr-default
- Chrome for Developers, 2026-06-17 — developer.chrome.com/blog/new-in-webgpu-149-150
- freeCodeCamp / Ajay Kalal, 2025-08-25 — www.freecodecamp.org/news/how-webgl-and-threejs-power-interactive-online-stores
- London Dynamics, 2025-02-28 — londondynamics.com/blog/ecommerce/retailers-guide-to-3d-ecommerce
- Research and Markets 3D E-Commerce Global Market Report 2026 — www.researchandmarkets.com/reports/6226870/3d-e-commerce-global-market-report