Skip to content

Benchmark

MTEB(cmn, 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
14
Scores recorded
14
Sources
1
IQ weight
1.04

Score distribution

71.4%
Mean
70.7%
Median
6.7%
Std dev
64.2%
10th percentile
83.3%
90th percentile
62.8%
Min
62.8%84.2%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 1.04 for embedding models, measured over 13 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Qwen3 Embedding 8BPairClassification· MTEB Leaderboard data84.2%
2Qwen3 Embedding 4BPairClassification· MTEB Leaderboard data83.3%
3BGE multilingual gemma2PairClassification· MTEB Leaderboard data79.3%
4Qwen3 Embedding 0.6BPairClassification· MTEB Leaderboard data76.4%
5bge-large-zhPairClassification· MTEB Leaderboard data72.9%
6Jina Embeddings v5 Text SmallPairClassification· MTEB Leaderboard data70.7%
7Jina Embeddings v5 Omni SmallPairClassification· MTEB Leaderboard data70.7%
8Jina Embeddings v5 Omni NanoPairClassification· MTEB Leaderboard data69.5%
9Jina Embeddings v3PairClassification· MTEB Leaderboard data67.2%
10Multilingual E5 Large InstructPairClassification· MTEB Leaderboard data66.8%
11E5 Mistral 7B InstructPairClassification· MTEB Leaderboard data66.3%
12Text Embedding Ada 002PairClassification· MTEB Leaderboard data65.2%
13Multilingual E5 LargePairClassification· MTEB Leaderboard data64.2%
14Text2vec Base ChinesePairClassification· MTEB Leaderboard data62.8%