Skip to content

Benchmark

MMEB Image - Visual7W

Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.

Models measured
1
Scores recorded
1
Sources
1
IQ weight
No weight published

Score distribution

70.9%
Mean
70.9%
Median
0%
Std dev
70.9%
10th percentile
70.9%
90th percentile
70.9%
Min

Leaderboard

#ModelScore
1Qwen3-VL Embedding 8BVisual7W· MMEB (TIGER-Lab) data70.9%

Data credits: MMEB Leaderboard (TIGER-Lab), Apache-2.0 (licence)