WorldReasonBench - Video Quality: leaderboard
WorldReasonBench (2026) tests video generators as future world-state predictors: from an initial frame and an action instruction, the generated video must reach the correct final state through a plausible process. This board is the multi-dimensional quality score, Qwen3.5-27B ratings of reasoning quality, temporal consistency and visual aesthetics mapped to 0-100; higher is better.
Metric: Multi-dimensional quality S(v) (0-100): Qwen3.5-27B point-wise ratings of reasoning quality (weight 0.4), temporal consistency (0.3) and visual aesthetics (0.3), overall on the 80-case shared evaluation set of WorldReasonBench (image-plus-text to video: from an initial frame and an action instruction, the generator must produce a video reaching the correct future world state; one generated clip per case); higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 11 models tracked. Updated 7 Oct 2026.
Top models
| # | Model | Score | Overall rank |
|---|---|---|---|
| 1 | Seedance 2.0 | 59.4 | |
| 2 | Sora2 | 56.9 | |
| 3 | WorldReasonBench Kling (checkpoint unspecified) | 55.4 | |
| 4 | Veo 3.1 Fast | 54.8 | |
| 5 | Wan2.6 | 50.3 | |
| 6 | Cosmos-Predict2.5 | 30.5 | |
| 7 | Wan2.2-14B | 30 | |
| 8 | LTX2.3 | 28.1 | |
| 9 | HunyuanVideo-1.5 | 27 | |
| 10 | LongCat-Video | 25.3 | |
| 11 | UniVideo | 21.3 |
Interactive version: theaggregate.ai/benchmark?slug=worldreasonbench-video-quality · How It Works · Data refreshed daily, snapshot 2026-11-22.