MMBench EN-DEV v1.1Multimodal

MMBench EN-DEV v1.1 is the English development split of OpenLab's general multimodal benchmark. It contains more than 3,000 questions covering perception, reasoning, knowledge, computation, and other dimensions, with ChatGPT assisting construction and validation of the question-answer pairs.

Ranking Sources

#ModelScore

1

Qwen3.5-397B-A17BAlibaba

93.7%

2

Gemini 3 ProGoogle

93.7%

3

Qwen3.6-35B-A3BAlibaba

92.8%

4

Qwen3.6-27BAlibaba

92.3%

5

Qwen3.5-35B-A3BAlibaba

91.5%

6

Claude Opus 4.5Anthropic

89.2%

7

Claude Sonnet 4.5Anthropic

88.3%

8

GPT-5.2OpenAI

88.2%

8 models total

Back to benchmarks

MMBench EN-DEV v1.1Multimodal

Official website

MMBench EN-DEV v1.1 is the English development split of OpenLab's general multimodal benchmark. It contains more than 3,000 questions covering perception, reasoning, knowledge, computation, and other dimensions, with ChatGPT assisting construction and validation of the question-answer pairs.

Ranking Sources

#ModelScore

1

Qwen3.5-397B-A17BAlibaba

93.7%

2

Gemini 3 ProGoogle

93.7%

3

Qwen3.6-35B-A3BAlibaba

92.8%

4