animation2code benchmark
For best compatibility, please view this dashboard in a Chrome browser.
← back to model results

ground truth

Nice spinny stuff

model outputs

Qwen3-VL-8B-Instruct (no output)
A T
GPT-5.4
A 0.91T 0.33
LLaMA 4 Scout
A 0.54T 0.33
1<div class="container">
2<div class="verypoop lol"></div>
3</div>