Benchmark
BLINK - Multi-view_Reasoning
Vision benchmarks measure how well models understand images. Scores are percentages.
All vision benchmarksScores from EvalAI data.
- Models measured
- 4
- Scores recorded
- 4
- Sources
- 1
- IQ weight
- No weight published
Score distribution
- 49.8%
- Mean
- 53.4%
- Median
- 6.3%
- Std dev
- 41.4%
- 10th percentile
- 57.9%
- 90th percentile
- 41.4%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | Claude 3 OpusMulti-view_Reasoning· EvalAI data | 57.9% |
| 2 | Qwen-VL MaxMulti-view_Reasoning· EvalAI data | 53.4% |
| 3 | LLaVA 1.6 34BMulti-view_Reasoning· EvalAI data | 46.6% |
| 4 | Gemini ProMulti-view_Reasoning· EvalAI data | 41.4% |