Bipolar transistors can have various base-regions width and various distances from the base-bond-wire to the useful base located between emitter and collector; this distance, and the aspect ratio of the path, and the resistance per cube of the path material, determine the rbb' which is a huge factor in setting the Boltzmann/Johnson/Nyquist random noise floor.
That same base region sets the transit time between emitter and collector, which sets Ftau (also called Falpha, and scaled by 1/beta becomes Fbeta); thus the maximum useful frequency for untuned amplifiers is set by the base region; oscillators may work a bit faster than Ftau because of energy resonances and thus more optimal energy movement.
Current specs and Voltage specs have already been mentioned.
Thermal resistance will vary dramatically, determined by die size and how heat can exit the bipolar die; if only bond wires to remove heat, expect 500 degree Centigrade per watt, or more. If a heavy copper slab is under the die, and your package provides access to the underside of that copper slab in a heat removal scheme, you may have 1 or 2 degree Centigrate per watt.
Then there are the high frequency junction capacitances.
Also the huge logarithmic range of operation, from picoAmps to milliAmps, may have a slope of 1 or a slope of 2, or some slope between 1 and 2, determined by various nuances of bipolar construction I have never needed to explore. Just be aware the 0.058 volts of delta_Vbase per 10:1 change in current ... may not be an accurate number.
The details of the exponential behavior (0.058 volts per 10:1) is crucial in defining the Taylor Series model of bipolar distortion, useful in IP2 and IP3 predictions. As briefly mentioned in the prior paragraph, different part numbers may have different logarithmic values, which will change the IP2 and IP3
predictions (and the actual measured distortion properties).