The work aims to coordinate the relationship between static and dynamic customer demands in view of the changes in the customer demands for the fresh commodity distribution, and plan the path reasonably and reduce the total logistics cost. Firstly, an integer programming model was established for minimizing the total logistics cost in consideration of the diverse temperature control intervals, random order request time, and dynamic customer demands quantity. Then, an improved ant colony algorithm based on Gaussian mixture clustering was designed to solve this model, and a dynamic demands processing strategy was proposed to re-optimize routes. Next, the effectiveness of the proposed algorithm was verified through comparison with particle swarm optimization, genetic algorithm and whale optimization algorithm. Finally, with the fresh commodity distribution network in Chongqing as an example, the operational indicators before and after optimization were compared and analyzed. Besides, the relationship between the value loss level of fresh commodity and the total logistics cost was explored. After optimization, the total logistics cost decreased by 22.35%, in which the penalty cost, the value loss, distribution cost and temperature control cost reduced by 39.84%, 61.84%, 29.80% and 57.00%, respectively. The proposed model, algorithm and dynamic demands processing strategy can reasonably plan routes and effectively reduce the total cost, which provides a methodological reference for fresh commodity distribution network optimization with dynamic demands.