DATA-VIZ · every model, same prompt

Find the most interesting dataset + visualize it

10 models built this from the identical prompt. Click any one to open the live build.

GLM 5.2 · pickopen →
Opus 4.8open →
GPT-5.5open →
Gemini 3.5open →
Grok 4.3open →
Qwen 3.7open →
DeepSeek V4open →
Mistral Largeopen →
Kimi K2.7open →
GPT-OSS 120bopen →
▸ The prompt every model got
Find the most genuinely interesting REAL public dataset you can think of — then build a single self-contained HTML page that makes it explorable and self-explanatory: an interactive visualization a curious person could lose ten minutes in.

YOU pick the dataset, and the choice is part of the test — pick something surprising and meaningful, NOT 'iris', 'titanic', or 'gapminder'. Inline a representative sample of the data in the file; vanilla JS + inline SVG or <canvas>, no charting libraries. Include a short note on what the dataset is and why you chose it. Return only the HTML.