MMMUval

AI benchmark results and model performance

Type: percentageGeneral

72.07%

64.50% - 80.60%

64.50% - Qwen 2 VL 2B

Scores Over Time

Individual benchmark scores plotted by date.

No scores available to display.

Organisation	Model	Reported	Top Score	Info	Self Reported	Source
Qwen	Qwen 2 VL 2B	-	64.50%	inferred family alias from qwen2-vl-72b (score=0.4083; benches=15)	Yes	Source
Qwen	Qwen 2 VL 72B	-	64.50%	-	Yes	Source
Qwen	Qwen 3 VL 235B A22B Instruct	-	78.70%	-	Yes	-
Qwen	Qwen 3 VL 235B A22B Thinking	-	80.60%	-	Yes	-