One little dog.
Different models.

An AI coding benchmark you can watch

PomBench compares how language models turn the same creative brief into a working browser animation. Each model writes its own scene and rendering code. Watch the Pomeranian's movement, the puddle jump and the bone retrieval, then compare the results on a shared timeline.

Explore model-specific recordings, output tokens and known API costs. This is an observable animation programming task, with original model output and documented corrections. It is not a text-to-video generator or an overall ranking of AI intelligence.