One prompt, ten runs
Eight models built a fractal explorer from the same request, some at more than one effort level. Open any of them to try its controls, transitions and performance on your own device.
Sonnet 5.5 low finished first, in 3m 25s. Opus 5.5 max ran longest and wrote the most, 587k output tokens.
The prompt, given to every model
I want you to build a website that allows the user to explore different fractals. There should be cool transitions between them, a themed ui, and lots of buttons and sliders to try out. It needs to be efficient enough to run in the browser on lower end devices, but take advantage of when there's more power available to run the page.
The runs
Build time is how long each model took to finish. Output counts the tokens it wrote, reasoning included. Total adds the input it read on every call, most of it cached.
- Build time
- Output tokens
- Total tokens