Supercell 2x2x2 of NiO (Space group: C2/m, 32 symmetry operations), magnetic moments preserved
Parsing. pymatgen CifParser reads the _atom_site_moment_ loop into site_properties["magmom"]: +2 μB on Ni0–Ni7, −2 μB on Ni8–Ni15, 0 on O. Net moment 0.00. It emits one benign warning ("No magnetic symmetry detected, using primitive symmetry") — moments survive anyway. Worth checking whether the visualizer path shows the same behavior.
Structure. 32 sites, density 6.85 g/cm³ (experiment 6.67), cell is a valid fcc supercell (basis {4a₃, 2a₁, 2a₂} of the fcc primitive cell, 16 primitive cells). All 16 fcc Ni positions occupied. Every Ni has 12 nearest-neighbor images at 2.885–3.011 Å — the split is the real type-II rhombohedral distortion, not noise. One counting trap: because b and c are only 2× the fcc primitive vector, the 12 NN images collapse onto 9 distinct site partners per Ni, so a neighbor-count that dedupes to site pairs will report 9. Both numbers are correct; they answer different questions. Each Ni is 6-coordinate in O at 2.052–2.116 Å.
Ordering. Textbook type-II AFM: per site, 6 NN images parallel (in the ferromagnetic {111} plane) + 6 antiparallel (adjacent planes), and all 48 second-shell pairs (4.10–4.13 Å) antiparallel. That is exactly the NiO signature, so the magmoms are not just preserved, they encode the correct magnetic structure.
One nuance. The moments are written along the crystal c axis, which in this setting is the cubic ⟨110⟩ direction — in the ferromagnetic (111) plane, but not the experimental NiO easy axis (spins prefer ⟨112⟩ within the plane). Irrelevant for a render/round-trip test; worth knowing if this file ever seeds a DFT calculation and you want the real ground-state constraint. Your C2/m description checks out too: the atomic frame analyzes as C2/m at symprec 0.1 (P-1 at 0.01 because of the small distortion jitters).