How much of the animation is clutter?
Leaderboard
How much of each model's animations the grader would cut, averaged over the run. Lower is better.
Cost and time
What an animation costs and how long it takes, against how much of it is clutter. Down and to the left is better.
Clutter against cost
Clutter against time
Pareto frontier: nothing is both cheaper and cleaneron the frontierbeatenone model across its reasoning efforts
Per graphic
Average cost
Average time
Every animation
How it works
The task
Clutter over time
Every element is tracked through the animation. A mark sits on its element in each frame, so it moves and fades with it. The score is the clutter in each frame, averaged over the run.
The grader
The same two passes of Claude Haiku 5.5 as the still graphics, with Claude Sonnet 5.5 breaking ties. Its rules add motion: title cards, decorative movement and end cards that restate the chart are marked remove.
Help rank the next benchmarks
Clutter is what should not be there. The next benchmarks are about what should, and that takes people picking the better of two graphics. Leave your email and we will write when there is something to rank.