Companies engaged in agro-industry, such as rice seed companies, depend on an efficient distribution process because of the characteristics of rice seed products that are easily damaged and do not last long. The distribution and delivery of goods from the production plant to reach consumers must go through several local distributors in several areas (multi-level) such as distributor centers, retailers, and agents spread across several cities in East Java Province. Determining the distribution network will be more complex when a company produces more than one type of product (multiproduct). Based on previous research, the genetic algorithm (GA) has been proven to provide the best solution for various optimization and combinatorial problems. However, the application of classical GA has the drawback that it has not yet reached the optimum global point, so it needs to be hybridized using a variable neighborhood search (VNS) algorithm. VNS was chosen because it can find solutions globally and find solutions locally to cover the shortcomings of GA. Using hybridization of GA-VNS, the cost obtained is 32392960, as evidenced by the cost savings of 323190 compared to the classic GA of 32716150. GA-VNS takes relatively the same time as classic GA.