We perform the first fit to the anisotropic clustering of Sloan Digital Sky Survey III CMASS data release 10 galaxies on scales of ∼0.8–32 h −1 Mpc. A standard halo occupation distribution model evaluated near the best-fitting Planck Λ cold dark matter (ΛCDM) cosmology provides a good fit to the observed anisotropic clustering, and implies a normalization for the peculiar velocity field of M ∼ 2 × 1013 h −1 M⊙ haloes of fσ8(z = 0.57) = 0.450 ± 0.011. Since this constraint includes both quasi-linear and non-linear scales, it should severely constrain modified gravity models that enhance pairwise infall velocities on these scales. Though model dependent, our measurement represents a factor of 2.5 improvement in precision over the analysis of DR11 on large scales, fσ8(z = 0.57) = 0.447 ± 0.028, and is the tightest single constraint on the growth rate of cosmic structure to date. Our measurement is consistent with the Planck ΛCDM prediction of 0.480 ± 0.010 at the ∼1.9σ level. Assuming a halo mass function evaluated at the best-fitting Planck cosmology, we also find that 10 per cent of CMASS galaxies are satellites in haloes of mass M ∼ 6 × 1013 h −1 M⊙. While none of our tests and model generalizations indicate systematic errors due to an insufficiently detailed model of the galaxy–halo connection, the precision of these first results warrant further investigation into the modelling uncertainties and degeneracies with cosmological parameters.