Skip to content

Benchmark

VQAv2 (val)

Vision benchmarks measure how well models understand images. Scores are percentages.

Models measured
3
Scores recorded
3
Sources
1
IQ weight
No weight published

Score distribution

68.3%
Mean
71%
Median
4.2%
Std dev
62.4%
10th percentile
71.6%
90th percentile
62.4%
Min
62.4%71.6%
The band covers the middle 80% of scores. The line marks the median.

Leaderboard

#ModelScore
1Gemma 3 12B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim71.6%
2Gemma 3 27B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim71%
3Gemma 3 4B IT· LLM Stats data· Reported by the model's developer. The source relaying it did not verify it. claim62.4%

Data credits: Data by LLM Stats