WBench - Physical Plausibility: leaderboard

The physical plausibility score of WBench (2026), a multi-turn benchmark for interactive video world models: text-driven, camera-controlled and action-conditioned video world models extend a scene turn by turn under their own controls. The score is the mean of causal fidelity and visual plausibility, on a 0-100 scale; higher is better.

Metric: Score (0-100): the mean of causal fidelity and visual plausibility, on the 158 navigation cases, multi-turn generation, each model under its own text, camera or action control; higher is better. Source: arxiv.org. Saturation forecast: Rough model projection: around 2026. 20 models tracked. Updated 7 Oct 2026.

Top models

#ModelScoreOverall rank
1Wan2.771.8
2LingBot-World71.2
3Kling 3.069.3
4LongCat-Video68.9
5Seedance 1.568.3
6Cosmos 2.567.4
7HY-Video 1.567.3
8Fantasy-World66.8
9HY-World 1.566.3
10Genie 365.7
11YUME 1.565.2
12InSpatio-World65.2
13LTX-2.364.8
14Happy Oyster63.5
15HY-GameCraft62.4

Interactive version: theaggregate.ai/benchmark?slug=wbench-physical-plausibility · How It Works · Data refreshed daily, snapshot 2026-11-21.