Benchmark
MTEB(cmn, v1) - Classification
Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.
- Models measured
- 13
- Scores recorded
- 13
- Sources
- 1
- IQ weight
- 0.91
Score distribution
- 63.1%
- Mean
- 66.1%
- Median
- 11.9%
- Std dev
- 43%
- 10th percentile
- 75.6%
- 90th percentile
- 43%
- Min
This benchmark has a weight of 0.91 for embedding models, measured over 13 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 | Qwen3 Embedding 8BClassification· MTEB Leaderboard data | 77.1% |
| 2 | Qwen3 Embedding 4BClassification· MTEB Leaderboard data | 75.6% |
| 3 | E5 Mistral 7B InstructClassification· MTEB Leaderboard data | 73% |
| 4 | Qwen3 Embedding 0.6BClassification· MTEB Leaderboard data | 71.5% |
| 5 | Multilingual E5 Large InstructClassification· MTEB Leaderboard data | 70.8% |
| 6 | Multilingual E5 LargeClassification· MTEB Leaderboard data | 69.8% |
| 7 | Paraphrase Multilingual MPNet v2Classification· MTEB Leaderboard data | 66.1% |
| 8 | Jina Embeddings v5 Text SmallClassification· MTEB Leaderboard data | 63.3% |
| 9 | Jina Embeddings v5 Omni SmallClassification· MTEB Leaderboard data | 63.3% |
| 10 | Jina Embeddings v5 Omni NanoClassification· MTEB Leaderboard data | 60.5% |
| 11 | All Mpnet Base V2Classification· MTEB Leaderboard data | 43.3% |
| 12 | All MiniLM L12 V2Classification· MTEB Leaderboard data | 43% |
| 13 | All MiniLM L6 V2Classification· MTEB Leaderboard data | 43% |