Benchmark
VN-MTEB (vie, v1) - Retrieval
Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.
- Models measured
- 8
- Scores recorded
- 8
- Sources
- 1
- IQ weight
- 1.42
Score distribution
- 31.1%
- Mean
- 39.8%
- Median
- 13.1%
- Std dev
- 14.1%
- 10th percentile
- 41.7%
- 90th percentile
- 9.6%
- Min
This benchmark has a weight of 1.42 for embedding models, measured over 8 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 | Jina Embeddings v5 Text SmallRetrieval· MTEB Extra Text Boards data | 44.4% |
| 2 | E5 Mistral 7B InstructRetrieval· MTEB Extra Text Boards data | 41.7% |
| 3 | Multilingual E5 Large InstructRetrieval· MTEB Extra Text Boards data | 40.9% |
| 4 | BGE M3Retrieval· MTEB Extra Text Boards data | 39.8% |
| 5 | Multilingual E5 LargeRetrieval· MTEB Extra Text Boards data | 37.7% |
| 6 | BGE multilingual gemma2Retrieval· MTEB Extra Text Boards data | 20.5% |
| 7 | All MiniLM L12 V2Retrieval· MTEB Extra Text Boards data | 14.1% |
| 8 | All MiniLM L6 V2Retrieval· MTEB Extra Text Boards data | 9.6% |