Benchmark
BLINK - Counting
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
- 59.6%
- Mean
- 65%
- Median
- 7.6%
- Std dev
- 49.2%
- 10th percentile
- 68.3%
- 90th percentile
- 49.2%
- Min
Leaderboard
| # | Model | Score |
|---|---|---|
| 1 | LLaVA 1.6 34BCounting· EvalAI data | 68.3% |
| 2 | Gemini ProCounting· EvalAI data | 65% |
| 3 | Qwen-VL MaxCounting· EvalAI data | 55.8% |
| 4 | Claude 3 OpusCounting· EvalAI data | 49.2% |