Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

M3²Bench

Benchmarks

Task NameDataset NameSOTA ResultTrend
Multimodal ReasoningM3-Bench web
ME Score46.4
24
Multimodal ReasoningM3-Bench robot
ME Score35.7
24
Text-to-CodeM3²Bench
Execution Rate (%)91.4
14
Diagram EditingM3²Bench
Execution Rate86.5
14
Diagram-to-CodeM3²Bench
Execution Rate94.3
14
Video Question AnsweringM3-Bench web
ME55.3
9
Video Question AnsweringM3-Bench robot
ME Score43.7
9
Long-video understandingM3-Bench web
MDR55.4
8
Long-video understandingM3-Bench robot
MDR47.8
8
Showing 9 of 9 rows