In general, the differences in the RepRaps are mostly about number of parts and cost of said parts. So if you can afford it, the most recent RepRap iterations (Prusa Mendel, etc) will give a complex build with good quality results.
If you want something that would be simpler to build, or cheaper, then the RepRap Wallace, or the Printrbot, would be similar builds, with reasonable quality results, and fewer parts with reduced complexity relative to the Prusa.
The Huxley is an alternative path to a simpler, smaller build prior to the Printrbot/Wallace group, so I don’t think I would recommend it at the moment - either build a full size Prusa, or a smaller Printrbot/Wallace if you need a smaller or cheaper bot.
The MendelMax is basically the same as the Prusa Mendel, but easier to build and more stable. It uses extruded aluminum parts, however, and is not self-replicating, so possibly not as much of a learning/teaching opportunity.
The RoStock is an interesting machine with similar properties to the Mendel’s in terms of being constructed from printed parts and vitamins, but is a newer design - might be fun and more interesting to kids that like robotics, but there is less support and experience available out there, and the mechanics of its operation are harder to explain and demonstrate, while perhaps being more interesting and offering more opportunity for experimentation (i.e. you are not dealing with a simple grid anymore, where a simple instruction to spin one motor causes a simple motion in one dimension, as with a Prusa).
Most of the other printers out there involve laser-cut parts for assembly (more akin to building a model plane or the like) and are not self-reproducing and thus not really RepRaps. (Note, this is also true of the non-printed Printrbot LC, etc.)
I would personally say to go for the Printrbot/Wallace - as a smaller, simpler bot to build, it would be more feasible for you to do multiple builds, i.e. have each new class build a new one each year, so that it is not merely the first class that gets the experience of building one, and it would be more within reach for the kids to feasibly replicate one for themselves.
As to deciding between the two, I’m not sure if there are sources out there selling Wallace kits or not, so I guess that might depend on whether you want the kids to have the experience of trying to find and source parts, in which case you can go with the Wallace, or merely of building the kit, in which case you can go with the Printrbot.
Hope that helps!