When modelling gravitational waves (GWs) from compact binaries, numerical relativity is well suited to modelling binaries with mass ratios q ~<= 15. At the other end of the mass ratio spectrum, linear black hole perturbation theory is ideal for modelling extreme mass ratio binaries with q > 10^4. In this talk I will show how we can bridge the gap between these two extremes and model intermediate mass ratio binaries using second-order self-force theory. I will discuss results from our recent Letter https://arxiv.org/abs/2107.01298 where we find remarkable agreement between GW fluxes computed from second-order self-force and numerical relativity at q=10 and smaller for quasi-circular binaries. I will also show new results comparing second-order self-force and numerical relativity waveforms.