Simple support unless you will be able to put significant torque into the supporting beams AND they have the required stiffness to get redistribution and stay serviceable.
You will never be wrong to assume simple support for mid-span reinforcing. Alternatively, given continuity on the ends, you _could_ design a span as two cantilevers, if you really want. The convention is somewhere between, and is based on strength and compatibility of the elements.
I jokingly tell people to "design for M=wl^2/8, and put half the steel in the top and half in the bottom, continuous throughout, develop the top at the ends and the bottom at midspan, and you're good to go." It's not quite so simple, but it gets you close without being excessive. Theoretically, in a continuous beam, once you know how much bottom mid-span reinforcement is required for your simple span, you could nearly divide it in any proportion between top end and bottom mid-span.