We model the dynamical structure of M87 (NGC4486) using high spatial resolution long-slit observations of stellar light in the central regions, two-dimensional stellar light kinematics out to half of the effective radius, and globular cluster velocities out to eight effective radii. We simultaneously fit for four parameters: black hole mass, dark halo core radius, dark halo circular velocity, and stellar mass-to-light (M/L) ratio. We find a black hole mass of 6.4(±0.5) × 10⁹ M _☉ (the uncertainty is 68% confidence marginalized over the other parameters). The stellar M/L_V = 6.3 ± 0.8. The best-fit dark halo core radius is 14 ± 2 kpc, assuming a cored logarithmic potential. The best-fit dark halo circular velocity is 715 ± 15 km s^–1. Our black hole mass is over a factor of 2 larger than previous stellar dynamical measures, and our derived stellar M/L ratio is two times lower than previous dynamical measures. When we do not include a dark halo, we measure a black hole mass and stellar M/L ratio that is consistent with previous measures, implying that the major difference is in the model assumptions. The stellar M/L ratio from our models is very similar to that derived from stellar population models of M87. The reason for the difference in the black hole mass is because we allow the M/L ratio to change with radius. The dark halo is degenerate with the stellar M/L ratio, which is subsequently degenerate with the black hole mass. We argue that dynamical models of galaxies that do not include the contribution from a dark halo may produce a biased result for the black hole mass. This bias is especially large for a galaxy with a shallow light profile such as M87, and may not be as severe in galaxies with steeper light profiles unless they have a large stellar population change with radius.