•A new approach for designing a green UHPC is proposed.•Two mathematical models based on DOMD and GA-ANN are built.•GA-ANN exhibits higher prediction accuracy compared with DOMD.•The GA-ANN is employed for mix design of UHPC successfully.•A green UHPC with advanced properties and dense matrix is developed. This paper presents a new approach for designing a green Ultra-High Performance Concrete (UHPC) with advanced particles packing meso-structure. To achieve this goal, two mathematical models based on D-Optimal Mixture Design (DOMD) and Genetic Algorithm based Artificial Neural Network (GA-ANN) are firstly established. After detailed comparison, it is demonstrated that the GA-ANN has higher accuracy and practicability compared to DOMD. To experimentally verify that the GA-ANN approach is suitable to be utilized in producing green UHPC, some UHPC specimens are developed based on this model, and their properties (workability, mechanical properties, pore structures and ecological characteristics) are investigated. The obtained experimental results show that the optimized UHPC has higher wet packing density and lower porosity, leading to excellent mechanical properties. Moreover, based on economic and environmental evaluation, the newly developed UHPC exhibits lower cost and environmental impact, something that makes it a sustainable construction product.