Benchmark
Arena ELO Vision Style-Controlled
People compare model answers and choose which they prefer. Higher scores are better.
All human preference benchmarksScores from LMArena data.
- Models measured
- 147
- Scores recorded
- 168
- Sources
- 1
- IQ weight
- 1.56
Score distribution
- 1,187.6
- Mean
- 1,215.3
- Median
- 96.7
- Std dev
- 1,025.4
- 10th percentile
- 1,287.8
- 90th percentile
- 881.3
- Min
This benchmark has a weight of 1.56 for chat models, measured over 147 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 |
|---|---|---|
| 151 | Pixtral 12B2409· LMArena data | 1,025.4 |
| 152 | InternVL2 26B· LMArena data | 1,025.2 |
| 153 | Nova Lite 1.0· LMArena data | 1,019.3 |
| 154 | Nova Pro 1.0· LMArena data | 1,019 |
| 155 | Claude 3 Sonnet20240229· LMArena data | 1,017.1 |
| 156 | Yi Vision· LMArena data | 1,003 |
| 157 | Aya Vision 32B· LMArena data | 1,000.8 |
| 158 | Claude 3 Haiku20240307· LMArena data | 1,000.4 |
| 159 | Molmo 7B-D0924· LMArena data | 994.5 |
| 160 | Llama 3.2 11B Vision Instruct· LMArena data | 993.1 |
| 161 | nvila-internal-15b-v1· LMArena data | 985.9 |
| 162 | LLaVA-OneVision Qwen2 72B OV· LMArena data | 978.6 |
| 163 | LLaVA 1.6 34B· LMArena data | 965 |
| 164 | CogVLM2 Llama3 Chat 19B· LMArena data | 964.2 |
| 165 | MiniCPM-V 2.6· LMArena data | 963 |
| 166 | InternVL2 4B· LMArena data | 958.7 |
| 167 | Phi-3.5 Vision Instruct· LMArena data | 919.7 |
| 168 | Phi-3 Vision 128K Instruct· LMArena data | 881.3 |
Showing 151–168 of 168 results.
Data credits: LMArena, CC BY 4.0 (licence)