I will assume that you have a power supply high enough to have 5 transistors stacked (one current source, two for cascoded input and two for cascoded load). If you want to reduce your UGB you need to add more output impedance or increase the transconductance of your input stage. You can achieve this in several ways, increase the length of the cascode transistors to increase their gds, try to assure that they are in saturation for all your input range. Also you can try to low a little your tail current, this will inherently increase the gm but will decrease the Vdsat=(vgs-vth) so you need to be careful to avoid enter in triode region. The limit of achieveable gain for a particular current depends on vdsat, the recommended level is around 150mV but this will depend in the technology used and voltage headroom.