Skip to content

Benchmark

MLEB - Consumer Contracts QA

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
1.04

Score distribution

85.2%
Mean
85.6%
Median
3.6%
Std dev
83.3%
10th percentile
88.4%
90th percentile
74.2%
Min
74.2%89%
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 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-4Consumer Contracts QA· MLEB data89%
2voyage-4-largeConsumer Contracts QA· MLEB data88.8%
3voyage-4-liteConsumer Contracts QA· MLEB data88.4%
4Qwen3 Embedding 8BConsumer Contracts QA· MLEB data88.4%
5Qwen3 Embedding 4BConsumer Contracts QA· MLEB data87.9%
6Gemini Embedding 001Consumer Contracts QA· MLEB data87.6%
7Jina Embeddings v5 Text SmallConsumer Contracts QA· MLEB data86.6%
8snowflake-arctic-embed-l-v2.0Consumer Contracts QA· MLEB data85.7%
9Text Embedding 3 LargeConsumer Contracts QA· MLEB data85.6%
10Text Embedding 3 SmallConsumer Contracts QA· MLEB data85.5%
11Text Embedding Ada 002Consumer Contracts QA· MLEB data85.4%
12Jina Embeddings v4Consumer Contracts QA· MLEB data85.2%
13Jina Embeddings v5 Text NanoConsumer Contracts QA· MLEB data85%
14snowflake-arctic-embed-m-v2.0Consumer Contracts QA· MLEB data84.7%
15Qwen3 Embedding 0.6BConsumer Contracts QA· MLEB data84.5%
16BGE M3Consumer Contracts QA· MLEB data83.3%
17Granite Embedding Small English R2Consumer Contracts QA· MLEB data78.4%
18Multilingual E5 Large InstructConsumer Contracts QA· MLEB data74.2%

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