Skip to content

Benchmark

Agentset Embedding nDCG@10 - DBPedia

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

Models measured
15
Scores recorded
15
Sources
1
IQ weight
0.62

Score distribution

78.5%
Mean
78.9%
Median
2.2%
Std dev
73.7%
10th percentile
80.7%
90th percentile
73%
Min
73%81.1%
The band covers the middle 80% of scores. The line marks the median.

This benchmark has a weight of 0.62 for embedding models, measured over 15 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-4DBPedia· Agentset81.1%
2Text Embedding 3 SmallDBPedia· Agentset80.7%
3Jina Embeddings v5 Text SmallDBPedia· Agentset80.5%
4Qwen3 Embedding 8BDBPedia· Agentset79.7%
5Text Embedding 3 LargeDBPedia· Agentset79.5%
6Gemini Embedding 2DBPedia· Agentset79.2%
7voyage-3-largeDBPedia· Agentset79%
8Jina Embeddings v3DBPedia· Agentset78.9%
9Qwen3 Embedding 4BDBPedia· Agentset78.7%
10voyage-3.5-liteDBPedia· Agentset78.7%
11BGE M3DBPedia· Agentset78.5%
12Embed Multilingual V3.0DBPedia· Agentset78.3%
13voyage-3.5DBPedia· Agentset78.2%
14Text Embedding 004DBPedia· Agentset73.7%
15Qwen3 Embedding 0.6BDBPedia· Agentset73%