Notes. The top block shows parameter constraints using the default conservative Planck lensing likelihood, alone and in combination with the DES lensing likelihood, the DES combined lensing and galaxy clustering likelihood, the Planck CMB acoustic scale, as well as the full Planck CMB power spectra. The middle block shows results for lensing alone when changing the lensing multipole range, using only temperature reconstruction (TT) rather than the minimum-variance combination with polarization (MV), or changing the binning scheme. In this context, “conservative” and “aggressive” refer to the individual bin boundaries listed in the upper and lower parts, respectively, of Table 1, so that, for instance, “MV aggressive 8 ≤ L ≤ 425” uses the first nine bins in the lower part of that table. The “CompSep mask” row shows results when constructing the lensing mask using the final common mask from Planck Collaboration IV (2020), rather than the earlier SMICA mask used by default throughout this paper. The lower block gives results from the default conservative (or aggressive) lensing likelihood when varying assumptions. The best-fit row shows the result of using a fixed ΛCDM fit to Planck TT,TE,EE+lowE for the CMB power spectra (as in the 2015 analysis) rather than marginalizing out the theoretical CMB spectra. The two “DES prior” rows (following DES Collaboration 2018b) use flat priors on 0.1 <  Ω_m <  0.9, 0.03 <  Ω_b <  0.07, 0.87 <  n_s <  1.07, 0.55 <  h <  0.91, 0.5 <  10⁹A_s <  5, and, when varying neutrino mass, 0.05 eV <  ∑m_ν <  1 eV (which is then unconstrained over this interval). All other results use flat priors on 0.001 <  Ω_ch² <  0.99, 0.5 <  θ_MC <  10, 1.61 <  log(10¹⁰A_s) < 3.91 and (except when combined with Planck CMB power spectra) our default lensing priors of 0.4 <  h <  1, n_s = 0.96 ± 0.02, Ω_bh² = 0.0222 ± 0.0005; when varying the neutrino mass, the flat prior is ∑m_ν <  5 eV with three degenerate neutrinos. Note these prior sensitivities have no impact on joint constraints with the CMB power spectra, where the lensing likelihood can be calculated self-consistently without additional priors. The small sensitivity to nonlinear modeling is demonstrated by using the Takahashi et al. (2012) variant of the halofit nonlinear model (Smith et al. 2003) rather than the default HMcode (Mead et al. 2016), and by comparison to the linear theory result when the nonlinear corrections are entirely neglected. All limits in this table are 68% intervals, and H₀ is in units of km s⁻¹ Mpc⁻¹.