Skip to content

Benchmark

MLEB - GDPR Holdings Retrieval

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

Models measured
18
Scores recorded
18
Sources
1
IQ weight
0.67

Score distribution

87.3%
Mean
87%
Median
2.9%
Std dev
83.8%
10th percentile
91.2%
90th percentile
82.9%
Min
82.9%93.8%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.67 for embedding models, measured over 18 models. The weight is how much it counts toward llminfo IQ. We measure it from the data, not from the raw scores.

Leaderboard

#ModelScore
1Gemini Embedding 001GDPR Holdings Retrieval· MLEB data93.8%
2voyage-4GDPR Holdings Retrieval· MLEB data91.6%
3voyage-4-largeGDPR Holdings Retrieval· MLEB data91.2%
4voyage-4-liteGDPR Holdings Retrieval· MLEB data90.4%
5Text Embedding 3 LargeGDPR Holdings Retrieval· MLEB data89.6%
6snowflake-arctic-embed-l-v2.0GDPR Holdings Retrieval· MLEB data87.8%
7snowflake-arctic-embed-m-v2.0GDPR Holdings Retrieval· MLEB data87.3%
8Jina Embeddings v5 Text NanoGDPR Holdings Retrieval· MLEB data87.2%
9Jina Embeddings v5 Text SmallGDPR Holdings Retrieval· MLEB data87%
10Text Embedding Ada 002GDPR Holdings Retrieval· MLEB data86.8%
11Qwen3 Embedding 4BGDPR Holdings Retrieval· MLEB data86.5%
12Jina Embeddings v4GDPR Holdings Retrieval· MLEB data86.1%
13Qwen3 Embedding 8BGDPR Holdings Retrieval· MLEB data85.4%
14Text Embedding 3 SmallGDPR Holdings Retrieval· MLEB data85.4%
15Multilingual E5 Large InstructGDPR Holdings Retrieval· MLEB data84.4%
16Qwen3 Embedding 0.6BGDPR Holdings Retrieval· MLEB data83.8%
17BGE M3GDPR Holdings Retrieval· MLEB data83.5%
18Granite Embedding Small English R2GDPR Holdings Retrieval· MLEB data82.9%

Data credits: MLEB (Isaacus), MIT (licence)