Benchmark
VN-MTEB (vie, v1) - STS
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
- 1.07
Score distribution
- 74.5%
- Mean
- 80.7%
- Median
- 9.9%
- Std dev
- 62.3%
- 10th percentile
- 82.9%
- 90th percentile
- 58.3%
- Min
This benchmark has a weight of 1.07 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 SmallSTS· MTEB Extra Text Boards data | 85.9% |
| 2 | Multilingual E5 Large InstructSTS· MTEB Extra Text Boards data | 82.9% |
| 3 | E5 Mistral 7B InstructSTS· MTEB Extra Text Boards data | 81.2% |
| 4 | Multilingual E5 LargeSTS· MTEB Extra Text Boards data | 80.7% |
| 5 | BGE M3STS· MTEB Extra Text Boards data | 78.8% |
| 6 | BGE multilingual gemma2STS· MTEB Extra Text Boards data | 66.1% |
| 7 | All MiniLM L12 V2STS· MTEB Extra Text Boards data | 62.3% |
| 8 | All MiniLM L6 V2STS· MTEB Extra Text Boards data | 58.3% |