Comment by frenchtoast8

They don’t benchmark different reasoning levels for Deepseek models.