The development of a universal influenza vaccine to elicit broad immune responses is essential in reducing disease burden and pandemic impact. In this study, the mosaic vaccine design strategy and genetic algorithms were utilized to optimize the seasonal influenza A virus (H1N1, H3N2) hemagglutinin (HA) and neuraminidase (NA) antigens, which also contain most potential T-cell epitopes. These mosaic immunogens were then expressed as virus-like particles (VLPs) using the baculovirus expression system. The immunogenicity and protection effectiveness of the mosaic VLPs were compared to the commercial quadrivalent inactivated influenza vaccine (QIV) in the mice model. Strong cross-reactive antibody responses were observed in mice following two doses of vaccination with the mosaic VLPs, with HI titers higher than 40 in 15 of 16 tested strains as opposed to limited cross HI antibody levels with QIV vaccination. After a single vaccination, mice also show a stronger level of cross-reactive antibody responses than the QIV. The QIV vaccinations only elicited NI antibodies to a small number of vaccine strains, and not even strong NI antibodies to its corresponding vaccine components. In contrast, the mosaic VLPs caused robust NI antibodies to all tested seasonal influenza virus vaccine strains. Here, we demonstrated the mosaic vaccines induces stronger cross-reactive antibodies and robust more T-cell responses compared to the QIV. The mosaic VLPs also provided protection against challenges with ancestral influenza A viruses of both H1 and H3 subtypes. These findings indicated that the mosaic VLPs were a promising strategy for developing a broad influenza vaccine in future.