Maltose and peptone were the best carbon and nitrogen sources for the production of destruxins from Metarhiziumanisopliae. With the addition of 0.1% (w/v) β‐alanine to the basal medium, the yields of cyclodepsipeptides DA and DB were 7.2 and 279 mg/L, respectively, which was 2‐fold higher than that of control experiment. Response surface methodology (RSM) was applied to optimize the compositions of maltose, peptone, β‐alanine, and glucose used in a shaker‐flask cultivation of M. anisopliae for the production of DA and DB. Estimated optimal compositions for the DA production were maltose 2.58%, peptone 0.72%, β‐alanine 0.02%, and glucose 0.55%. The predicted DA yield was 18.5 mg/L. On the other hand, the optimal compositions for DB production were maltose 2.51%, peptone 0.75%, β‐alanine 0.02%, and glucose 0.43%. A maximum DB yield of 232 mg/L was predicted. These were confirmed by cultivation experiments conducted at the optimized conditions for maximum destruxins production in a shaker‐flask. Furthermore, a modest high level of DA (49 mg/L) and DB (268 mg/L) yields were obtained by employing the response surface methodology optimized DB production medium in a no‐baffle, stirred‐tank fermentor.