Abstract: The multi-product newsvendor problem based on the conditional value-at-risk (CVaR) is studied quantitatively. Firstly, to gain maximum the sum of CVaR of products' revenue, a multi-product newsvendor model with budget constraint was proposed. The marginal budget CVaR was non-decreasing in the order quantity. A binary search algorithm was developed. Finally, the effects of budget and risk aversion were analyzed through numerical example. The research results show that the greater the budget and risk aversion are, the more categories of products are ordered by the retailer; and the expected revenue of retailer is decreasing in the risk aversion.