When "shape" is used in the context of gamma-ray bursts, it is typically used to refer to the jets emitted on opposite ends of an axis through the source. These jets contain huge amounts of energy, most in the form of photons so high in energy that they are classified as gamma-rays, some of the highest-frequency photons known.
The jets are typically emitted along an axis through the progenitor object (I'm not going to talk about the object itself, as the causes of GRBs are not yet fully understood). A quick search engine search (I prefer duckduckgo, but you might like Google better) can give you hundreds of pictures (mostly artists' impressions) of GRBs. Notice how thin the jets are. They look like long cylinders, but closer inspection shows that they aren't. They are actually cones (truncated ones, but cones nonetheless).
The reason that these jets spread out is relativistic beaming, which also affect the beam's luminosity. The angle at which the beams diverge is so slim - in fact, this pre-print says that one observed GRB had a jet emitted at an angle of one tenth of a radian! That's only 18 degrees. (Note: On page 21, this calls an angle of 3 degrees "typical" for some jets ).
So the answer to your question is that gamma-ray bursts are typically conical, although the jets are often nearly cylindrical. The first paper I mentioned also did an analysis of spherical models, but concluded that they did not fit observations.