Benchmark
ViDoRe(v3.1) - Public split
Vision benchmarks measure how well models understand images. Scores are percentages.
- Models measured
- 8
- Scores recorded
- 8
- Sources
- 1
- IQ weight
- 1.15
Score distribution
- 50.5%
- Mean
- 57%
- Median
- 10.9%
- Std dev
- 44.4%
- 10th percentile
- 58.8%
- 90th percentile
- 25.9%
- Min
This benchmark has a weight of 1.15 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 | Nemotron 3.5 Embed 8bPublic split· MTEB Multimodal Boards data docs | 61.5% |
| 2 | Nemotron 3.5 Embed 1bPublic split· MTEB Multimodal Boards data docs | 58.8% |
| 3 | Nemotron 3 Embed 1bPublic split· MTEB Multimodal Boards data docs | 57.8% |
| 4 | Qwen3-VL Embedding 8BPublic split· MTEB Multimodal Boards data docs | 57% |
| 5 | Llama Nemotron Embed VL 1b V2Public split· MTEB Multimodal Boards data docs | 52.6% |
| 6 | Jina Embeddings v5 Omni NanoPublic split· MTEB Multimodal Boards data docs | 45.7% |
| 7 | Jina Embeddings v5 Omni SmallPublic split· MTEB Multimodal Boards data docs | 44.4% |
| 8 | bge-large-enPublic split· MTEB Multimodal Boards data docs | 25.9% |