This research was conducted in the industry producing cutting tools with high variations such as cutters, drillers, and reamer. Products are made according to consumer demand with a hybrid job shop production flow. The problem faced by the company is that many jobs have passed due dates or are late, which is caused by incorrect production scheduling. The purpose of this study is to design a scheduling that can minimize the mean tardiness. To determine the scheduling the Firefly algorithm method is used which minimizes the makespan. Data processing is performed using 25 jobs as numerical examples from 783 jobs that will be scheduled from the results of sorting based on the level of delay. After scheduling, completion time, number of tardy jobs, and mean tardiness are calculated. Scheduling with the Firefly algorithm produces a makepan of 1395 minutes, so the number of tardy jobs decreases from 25 jobs to 3 jobs with each delay for 1 day, and mean tardiness decreases from 11.36 days to 0.12 days.