Don't take our word for it - play it.

We asked models across the capability ladder - from a tiny 8B you can run on a laptop to a closed frontier flagship - for a single self-contained index.html: a game, a 3D scene, a toy, a page, a calculator. Below is exactly what each one produced, the whole thing scaled to fit. Same prompt, same agent - the difference is the model.

Thumbnails are live, sandboxed, scaled to fit (no scrolling). Hit open to play one full size.