If you want evil bombs, we already have nukes.
https://en.wikipedia.org/wiki/Non-neutral_plasma
(see discussion of the "Brillouin Limit")
Here, one might exploit the reaction p(p,pi+)d (or, rather, it's antimatter equivalent, which makes a negative pion), at a center of mass energy above threshold for creating the charged pion (which has a mass of 139.579 MeV. This is wildly energy negative, but if one has already invested many GeV in making each antiproton that's presumably acceptable. The cross section for this reaction is only ~200 microbarns, but that will be many orders of magnitude higher than the ordinary fusion cross section.
The solid antimatter one would probably target as an end goal would be anti-(lithium hydride).
Didn't stop people then. And it won't stop sufficiently criminal governments today.
And we don't even have to reach for nukes. Dynamite is also a better choice for blowing things up than antimatter.
Now do the same for Gaza and anywhere other than WWII Warsaw. Carpet bombing isn’t necessary if you can aim with precision at the support infrastructure of occupied structures.
Carpet bombing is not meant for eliminating military targets - it's an act of state-sponsored terrorism to get a population to rise against its leaders. Only, it has been proven over and over again: That doesn't work and often results in the exact opposite, making them rally behind the flag against the barbarous enemy who would attack civilians. Which is why we stopped doing that. Mostly.
Only about 0.01% of the energy used to operate the particle collider creates antimatter, the vast majority of which is impossible to capture. All in all, the efficiency of the entire process - if you were to measure it in the e^2=(pc)^2+(mc^2)^2 sense - is probably on the order of 1e-9 or worse.
tl;dr: strong magnetic fields in a certain configuration since in a strong enough field anti-hydrogen acts like little bar magnets.
Not to mention the only way to create it is with energy (it doesn't exist on Earth), and we can only do so at terrible efficiencies. So even theoretically it's pretty bad.