Relative variations of the matter power spectrum (left panel) and bispectrum in the squeezed limit (k₁ = k₂ = k; k₃ = 3k_f) (right panel) with respect the reference simulation using MonofonIC (see Section 3.3). In the upper panels, we show the variations at the initial redshift, z = 49, and in the lower panels at z = 0. The codes are shown in different colours: 2LPTPNG (blue), LPICOLA (orange), FASTPM k_max = 1 (red), FastPM k_max = 2/3 (green), and FastPM k_max = 1/4 (purple). The vertical dashed lines show k = k_nyq × 1/4, and the vertical dash-dotted lines k = k_nyq × 2/3. The grey shaded regions represent 1% differences with respect to the reference. LPICOLA shows the largest variations at small k, due to a different realisation of the white noise than the reference simulation (see Section 3.2). Notably, FastPM with k_max = 1/4 shows a 80% deviation for $k>\frac{1}{4}{k}_{nyq}$ for the bispectrum at the initial conditions (see Section 4.3). 2LPTPNG and FastPM, with k_max = 1 and 2/3, show an excellent agreement with respect MonofonIC, with variations below ~0.1% in the power spectrum and bispectrum at all scales, for both the initial conditions and at z = 0.