AnimationBench - Distinctive Content: leaderboard
Metric: Distinctive Content: whether the video shows the uncommon, animation-characteristic elements the prompt asks for (%), animation-style image-to-video generation from the AnimationBench close-set suite (fixed source images and prompts, about 30 videos per dimension per model), scored by Qwen3-VL-Max yes/no question sets aggregated per video; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 8 models tracked.
Top models
| # | Model | Score |
|---|---|---|
| 1 | Veo 3.1 | 90 |
| 2 | Seedance 2.0 | 86.67 |
| 3 | Sora 2 Pro | 78.33 |
| 4 | Kling 2.6 | 75 |
| 5 | Seedance 1.0 Pro | 70 |
| 6 | Wan2.2-I2V-A14B | 61.67 |
| 7 | FramePack | 35.83 |
| 8 | HunyuanVideo-1.5 | 20.83 |
Interactive version: theaggregate.ai/benchmark?slug=animationbench-distinctive-content · How It Works · Data refreshed daily, snapshot 2026-10-07.