What Meshopt does
The extension provides a generic compression for binary data that is tailored to typical glTF content [1]. The focus is on fast unpacking: a 3D model should appear soon after download without the browser crunching for long.
Quantisation
Meshopt works cleanly with KHR_mesh_quantization by compressing already quantised data [1]. Quantisation stores positions, normals or UVs with fewer bits, for example as 16-bit integers instead of 32-bit floats [2]. For product views the loss of precision is usually invisible, while the data becomes much smaller.
Tool: gltfpack
The command-line tool gltfpack optimises glTF files for smaller downloads: it reorders data for the GPU cache, quantises, merges meshes and can compress with Meshopt [3]. It can also convert textures to KTX2 or WebP [3].
In the browser
In three.js the Meshopt decoder is registered with the GLTFLoader [4]. Files using this compression only open in viewers that support the extension, so check before passing the same file to third parties.
Meshopt or Draco?
Both shrink geometry. Draco reaches higher compression on some models; Meshopt is built for decoding speed. There is no blanket answer; measure on your own model.
From our library
All GLB files in our example configurators use Meshopt with quantisation, for instance the planetary gearhead, the beverage can and the window configurator (fictional products). The guide CAD to web 3D puts compression in context.