Benchmark
Agentset Embedding Win Rate
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
- 0.96
Score distribution
- 46.6%
- Mean
- 47%
- Median
- 8.5%
- Std dev
- 34.6%
- 10th percentile
- 57%
- 90th percentile
- 28.4%
- Min
This benchmark has a weight of 0.96 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 | Gemini Embedding 2· Agentset | 59.5% |
| 2 | voyage-4· Agentset | 57% |
| 3 | Text Embedding 3 Large· Agentset | 56.4% |
| 4 | Jina Embeddings v5 Text Small· Agentset | 54.7% |
| 5 | voyage-3-large· Agentset | 51.3% |
| 6 | Qwen3 Embedding 8B· Agentset | 48.8% |
| 7 | Embed Multilingual V3.0· Agentset | 48.4% |
| 8 | voyage-3.5· Agentset | 47% |
| 9 | Qwen3 Embedding 4B· Agentset | 44.6% |
| 10 | BGE M3· Agentset | 44.3% |
| 11 | voyage-3.5-lite· Agentset | 44.2% |
| 12 | Text Embedding 3 Small· Agentset | 43.9% |
| 13 | Qwen3 Embedding 0.6B· Agentset | 35.7% |
| 14 | Jina Embeddings v3· Agentset | 34.6% |
| 15 | Text Embedding 004· Agentset | 28.4% |