Skip to content

Benchmark

MLEB - ECHR 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
0.80

Score distribution

23.9%
Mean
21.6%
Median
10.5%
Std dev
13.4%
10th percentile
37.7%
90th percentile
10.4%
Min
10.4%46.4%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.80 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
1voyage-4ECHR Retrieval· MLEB data46.4%
2voyage-4-largeECHR Retrieval· MLEB data44.1%
3voyage-4-liteECHR Retrieval· MLEB data37.7%
4Qwen3 Embedding 4BECHR Retrieval· MLEB data32.2%
5Qwen3 Embedding 8BECHR Retrieval· MLEB data31%
6Qwen3 Embedding 0.6BECHR Retrieval· MLEB data28.7%
7Jina Embeddings v4ECHR Retrieval· MLEB data24.5%
8Jina Embeddings v5 Text NanoECHR Retrieval· MLEB data22.6%
9Jina Embeddings v5 Text SmallECHR Retrieval· MLEB data21.6%
10Text Embedding 3 LargeECHR Retrieval· MLEB data20.3%
11Text Embedding Ada 002ECHR Retrieval· MLEB data20%
12Text Embedding 3 SmallECHR Retrieval· MLEB data17.6%
13snowflake-arctic-embed-l-v2.0ECHR Retrieval· MLEB data17.5%
14BGE M3ECHR Retrieval· MLEB data17%
15Gemini Embedding 001ECHR Retrieval· MLEB data13.6%
16Granite Embedding Small English R2ECHR Retrieval· MLEB data13.4%
17snowflake-arctic-embed-m-v2.0ECHR Retrieval· MLEB data11.2%
18Multilingual E5 Large InstructECHR Retrieval· MLEB data10.4%

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