Benchmark
MLEB - Singaporean Judicial Keywords
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
- 0.81
Score distribution
- 81.2%
- Mean
- 84.4%
- Median
- 12.3%
- Std dev
- 63.4%
- 10th percentile
- 94.8%
- 90th percentile
- 57.5%
- Min
This benchmark has a weight of 0.81 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
| # | Model | Score |
|---|---|---|
| 1 | voyage-4-largeSingaporean Judicial Keywords· MLEB data | 96.8% |
| 2 | voyage-4Singaporean Judicial Keywords· MLEB data | 96.2% |
| 3 | voyage-4-liteSingaporean Judicial Keywords· MLEB data | 94.8% |
| 4 | Jina Embeddings v4Singaporean Judicial Keywords· MLEB data | 92.7% |
| 5 | Qwen3 Embedding 8BSingaporean Judicial Keywords· MLEB data | 92.5% |
| 6 | Qwen3 Embedding 4BSingaporean Judicial Keywords· MLEB data | 90.4% |
| 7 | Jina Embeddings v5 Text SmallSingaporean Judicial Keywords· MLEB data | 88.5% |
| 8 | Jina Embeddings v5 Text NanoSingaporean Judicial Keywords· MLEB data | 85.6% |
| 9 | Qwen3 Embedding 0.6BSingaporean Judicial Keywords· MLEB data | 84.4% |
| 10 | Text Embedding 3 LargeSingaporean Judicial Keywords· MLEB data | 83.5% |
| 11 | Gemini Embedding 001Singaporean Judicial Keywords· MLEB data | 78.3% |
| 12 | Text Embedding 3 SmallSingaporean Judicial Keywords· MLEB data | 77.7% |
| 13 | snowflake-arctic-embed-l-v2.0Singaporean Judicial Keywords· MLEB data | 77.2% |
| 14 | snowflake-arctic-embed-m-v2.0Singaporean Judicial Keywords· MLEB data | 73.8% |
| 15 | Text Embedding Ada 002Singaporean Judicial Keywords· MLEB data | 69.6% |
| 16 | Granite Embedding Small English R2Singaporean Judicial Keywords· MLEB data | 63.4% |
| 17 | BGE M3Singaporean Judicial Keywords· MLEB data | 58.3% |
| 18 | Multilingual E5 Large InstructSingaporean Judicial Keywords· MLEB data | 57.5% |
Data credits: MLEB (Isaacus), MIT (licence)