Skip to content

Benchmark

VBench-Long spatial relationship

Scores are percentages.

Models measured
9
Scores recorded
9
Sources
1
IQ weight
0.53

Score distribution

75.5%
Mean
74.3%
Median
8.6%
Std dev
66.2%
10th percentile
84.3%
90th percentile
65.4%
Min
65.4%93.7%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.53 for video-generation models, measured over 8 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Vidu Q12025-04-17· VBench data docs93.7%
2Veo 3· VBench data docs84.3%
3Kling V1.6· VBench data docs79.1%
4Video 01· VBench data docs75.5%
5Sora· VBench data docs74.3%
6Hunyuan Video2025-05-22· VBench data docs72.1%
7Mochi 1· VBench data docs69.2%
8Vidu Q4· VBench data docs66.2%
9LTX Video5s 768×512· VBench data docs65.4%

Data credits: VBench (Vchitect), MIT (licence)