Conventional breeding involves considerable genetic recombination. To maximize breeding efforts, it is necessary to increase seed yield. This study proposes to improve seed yield by using tetraploid bananas as part of a genetic improvement programme for cooking varieties. To this end, the female fertility (seeding rate and embryo sac maturity) of tetraploid banana plants was characterised. The pollen potential of tetraploids was then assessed at anthesis and at the end of meiosis, as well as during crosses with seminiferous diploids, leading to a phenotypic characterisation of the value of an offspring. This study highlighted the low seed production rate of tetraploid genitors. At best, 7% of seeds are produced per fruit, but 57% of the ovules per fruit have mature embryo sacs. The use of tetraploids as male parents, free from any source of sterility associated with the domestication process, did not generally improve the seed rate, except in the case of FHIA 21 (2930 seeds). Its pollen potential was used in a cross with Banksii 0623 (♀). This resulted in progeny with interesting vigour and phenotypic characteristics. This latter combination has the potential to transfer its “cooking” genetic structure.