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

ground truth

Only CSS: Codevember #5 Lightsaber

model outputs

Qwen3-VL-8B-Instruct (no output)
A T
GPT-5.4
A 0.83T 0.28
LLaMA 4 Scout
A 0.49T 0.20
1<div class="force_field"><div class="sword"><div class="grip"></div><div class="beam"></div></div></div>