I believe the title is pretty self-explanatory: I understand both roughly, but am unable to make a clear distinction.
We use a circuit formulation for which an LC circuit, being a resonator, is the cavity, while the two-level systems (qubit) is a JJC circuit (LC circuit with a Josephson junction instead of the inductor).
The Hamiltonian for the qubit is: \begin{equation} \hat{H} = \hbar \omega \Big(\hat{n} + \frac{1}{2} \Big) = \hbar \omega \Big( \mathbb{I}-\frac{\hat{\sigma_z}}{2} \Big) \equiv \hat{H}_{I}. \end{equation}
In the simplest case we consider only the qubit. In the circuit formulation, if the qubit is driven with some frequency, one models it by substituting $V \mapsto V_0 + cos(\omega_{d} t)$ and (after some math) finds a new term in the Hamiltonian: \begin{equation} \hat{H}_{II} = K cos(\omega _d t) \hat{\sigma_x}, \end{equation} where obviously $\hat{H}=\hat{H}_{I}+\hat{H}_{II}$ and $K$ is some constant.
So if one represents the qubit on the Bloch sphere, the driven Hamiltonian is composed of two terms: the one that is $\propto \hat{\sigma}_z$ is a "Larmor" precession about the $z$ axis (always present) while the one that is $\propto \hat{\sigma}_x$ is a precession about the $x$ axis, present only for driven qubits, and called Rabi oscillation; and the condition for which we pick a Bloch Sphere that is rotating in a way that cancels out the Rabi oscillations is called Rotating Wave Approximation (RWA).
(Is this correct up to this point ?)
Than, in the complete case where one considers the coupled atom-cavity, the full Hamiltonian is the Jaynes-Cummings Hamiltonian
\begin{equation} \hat{H}_{JC} = \hbar \omega _c \hat{a}^\dagger \hat{a} + \frac{\hbar \omega_a }{2} \hat{\sigma}_z + \hbar g ( \hat{a} \hat{\sigma}^+ + \hat{a}^\dagger \hat{\sigma}^-). \end{equation}
The three terms are related to the cavity, the atom and the coupling and $g$ is the coupling factor. The terms $a$ and $\sigma$ are creation/annihilation operators for the photon and raising/lowering operator for the electron in the atom: this is good, but I'm not able to work out the matrix form of $\hat{a} \hat{\sigma}^+ + \hat{a}^\dagger \hat{\sigma}^-$. I get the identity matrix $\mathbb{I}$, while it looks like it should be $\hat{\sigma}_x$. This would, in turn, give a case much similar to the previous one... only this time the oscillations would be called Vacuum Rabi Oscillations.
And at this point, I'm full of doubts. What's the physical difference? Why "vacuum" ? The qubit Rabi-oscillates without being actively driven: does this mean that coupling it to the cavity acts like some sort of passive driving, "vacuum" driving hence the name?