Bimanual sequential skills, such as typing, are essential for our modern lives. However, how the brain learns bimanual sequential skills is largely unknown. Bimanual sequential skill requires two essential leanings: bimanual posture control and representing motor sequence. We conducted an fMRI study to investigate the neural basis of bimanual sequential skills like typing, resembled in our ordinary movement. Thirty-five healthy right-handed volunteers practiced visually cued bimanual serial reaction time tasks, including the “mirror” and more difficult “parallel” modes of repeating fixed sequences or random movements to separately depict the neural substrates of bimanual posture control and those of sequence. The reaction time (RT) of random movements continuously declined, indicating learning of bimanual posture formation. The RT in the sequential condition declined more rapidly than in the random condition, confirming sequence learning. The random parallel conditions evoked a more prominent learning-related decrease of task-related activation in the left M1 and cerebellar vermis than the less difficult random mirror conditions. The left M1 showed learning-related enhancement of functional connectivity with the anterior cingulate cortex during the “parallel random” conditions compared with the “mirror random” conditions. Thus, the left M1, anterior cingulate cortex, and cerebellar vermis are related to learning bimanual posture control. The left M1 and cerebellar vermis also showed sequence-specific learning-related activity increments more prominent in the parallel mode than in the mirror mode. Thus, the left M1 and cerebellar vermis are critical in the bimanual motor learning network.