Skip to content

Benchmark

VN-MTEB (vie, v1) - PairClassification

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

Models measured
6
Scores recorded
6
Sources
1
IQ weight
1.55

Score distribution

78.1%
Mean
83.9%
Median
8.3%
Std dev
68.5%
10th percentile
84.6%
90th percentile
64.5%
Min
64.5%84.6%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.55 for embedding models, measured over 6 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1BGE M3PairClassification· MTEB Extra Text Boards data84.6%
2Jina Embeddings v5 Text SmallPairClassification· MTEB Extra Text Boards data84.5%
3Multilingual E5 LargePairClassification· MTEB Extra Text Boards data83.9%
4BGE multilingual gemma2PairClassification· MTEB Extra Text Boards data82.5%
5All MiniLM L12 V2PairClassification· MTEB Extra Text Boards data68.5%
6All MiniLM L6 V2PairClassification· MTEB Extra Text Boards data64.5%