Skip to content

Benchmark

GEditBench v2 - Visual Consistency

Models measured
12
Scores recorded
14
Sources
1
IQ weight
0.14

Score distribution

990.8
Mean
1,017
Median
79
Std dev
846
10th percentile
1,108
90th percentile
840
Min
8401,109
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.14 for image-generation models, measured over 12 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard