Tiny Online Tools
Find out why the model is slow before you start changing it
A 3D model that feels heavy on the web is rarely heavy for the reason people assume. Everyone looks at the file size, because that is the number the operating system shows. But file size is one of four things that decide how long a model takes to appear, and often not the dominant one. This analyzer measures all four and tells you which one is actually costing you.
The four phases of a model load
Download is bytes over the wire — the only phase that improves when you compress the file. Parse is JSON, buffer views and codec decode. Texture decode turns compressed images into RGBA pixels and builds mip chains. GPU upload pushes vertex and texture data across the bus.
On fibre, a 12 MB model with eight 4K textures spends under a second downloading and several seconds decoding textures. Compressing the GLB harder will not help that model at all. Halving the texture resolution will. The tool names the dominant phase for every connection speed so you can tell those two situations apart.
File size and GPU memory are not the same number
This is the single most common misunderstanding in web 3D. A 4096 × 4096 texture might ship as a 2 MB JPEG, but on the GPU it is 4096 × 4096 × 4 bytes of RGBA — about 67 MB, or roughly 89 MB once the mip chain exists. Eight of those and a mid-range phone runs out of memory regardless of how small the download was.
The analyzer reports GPU texture memory and vertex attribute memory separately from file size, and grades them separately, because they fail for different reasons and are fixed by different things.
Graded against three targets at once
Every metric is scored against desktop, mobile and WebXR budgets simultaneously, so the answer is not a yes or no but a where: this model is fine on desktop, marginal on mobile, and hopeless in a headset. Each target gets a letter grade, each metric gets a pass, warn or fail, and every failure produces one specific piece of advice naming the transform that fixes it.
Read-only by design
This tool never writes a file. Once you know what to fix, the 3D Web Optimizer applies the plan and re-grades the result. Everything happens locally — your models are never uploaded.