Benchmark
Agentset Embedding Win Rate - MSMARCO
Embedding benchmarks measure how well models find related items, for example when matching a search to relevant results. Scores are percentages. Higher is better.
- Models measured
- 15
- Scores recorded
- 15
- Sources
- 1
- IQ weight
- 0.52
Score distribution
- 44%
- Mean
- 45.6%
- Median
- 8.3%
- Std dev
- 30%
- 10th percentile
- 55.6%
- 90th percentile
- 27.5%
- Min
This benchmark has a weight of 0.52 for embedding models, measured over 15 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 8BMSMARCO· Agentset | 57.5% |
| 2 | voyage-3.5MSMARCO· Agentset | 55.6% |
| 3 | Qwen3 Embedding 4BMSMARCO· Agentset | 52.5% |
| 4 | Text Embedding 3 SmallMSMARCO· Agentset | 49.1% |
| 5 | Gemini Embedding 2MSMARCO· Agentset | 47.3% |
| 6 | Text Embedding 3 LargeMSMARCO· Agentset | 46.9% |
| 7 | Jina Embeddings v5 Text SmallMSMARCO· Agentset | 46.7% |
| 8 | Embed Multilingual V3.0MSMARCO· Agentset | 45.6% |
| 9 | voyage-4MSMARCO· Agentset | 45% |
| 10 | voyage-3-largeMSMARCO· Agentset | 41.7% |
| 11 | voyage-3.5-liteMSMARCO· Agentset | 40% |
| 12 | Jina Embeddings v3MSMARCO· Agentset | 38.8% |
| 13 | BGE M3MSMARCO· Agentset | 35.6% |
| 14 | Text Embedding 004MSMARCO· Agentset | 30% |
| 15 | Qwen3 Embedding 0.6BMSMARCO· Agentset | 27.5% |