Benchmark
MTEB(kor, v2) - Retrieval
Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages.
- Models measured
- 7
- Scores recorded
- 7
- Sources
- 1
- IQ weight
- 1.52
Score distribution
- 73.1%
- Mean
- 75.6%
- Median
- 5.4%
- Std dev
- 70.2%
- 10th percentile
- 77.1%
- 90th percentile
- 61.2%
- Min
This benchmark has a weight of 1.52 for embedding models, measured over 7 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 8BRetrieval· MTEB Leaderboard data | 78.4% |
| 2 | Qwen3 Embedding 4BRetrieval· MTEB Leaderboard data | 77.1% |
| 3 | Llama Nemotron Embed VL 1b V2Retrieval· MTEB Leaderboard data | 75.6% |
| 4 | Qwen3-VL Embedding 8BRetrieval· MTEB Leaderboard data | 75.6% |
| 5 | snowflake-arctic-embed-l-v2.0Retrieval· MTEB Leaderboard data | 74% |
| 6 | BGE multilingual gemma2Retrieval· MTEB Leaderboard data | 70.2% |
| 7 | Granite Embedding 278m MultilingualRetrieval· MTEB Leaderboard data | 61.2% |