Stay away from anything stinking fast (say >3GHz) until you're ready. Even for anything fT between <your fastest instrument's bandwidth> and there, at least try to get something that can read the presence of extremely high frequencies, like a diode RF probe. Otherwise, when your circuit does inevitably oscillate, you'll have little idea what's going on, aside from "spooky" behavior like shifting bias or distortion with fingers nearby (proximity effect).
Si JFETs are generally good for the 100s of MHz range. As transconductance devices, they don't really have a transition frequency as such. BJTs can be expressed as current mode devices, and therefore have hFE and fT. But for a true constant current output, you can simply keep increasing the load resistance, and power gain keeps going up (if not maximum power output, because you'll pretty quickly run out of voltage). You do have the direct tradeoff of band width (which is DC to BW for a wideband circuit, or center frequency +/- BW/2 for a tuned circuit), because drain circuit bandwidth is limited by load resistance versus drain capacitance. Likewise for gate/source circuit bandwidth tradeoff. At some point, bandwidth is too narrow to use (plus the real losses in gate and drain terminals) and you can't do much more with it.
Tim