Gemma 4 31B Instruct | BIG-Bench Extra Hard Original — micro average

Google (DeepMind)'s Gemma 4 31B Instruct evaluation results on BIG-Bench Extra Hard Original — micro average

Score: 74.4%

Reasoning level: thinking mode

Source

Gemma 4 31B Instruct (Registry No. 35002)

Google (DeepMind)

BIG-Bench Extra Hard Original — micro average