The crystal structure of (Be,¿)(V,Ti)₃O₆, a mineral discovered in the emerald deposit of Byrud, Norway, has been solved and refined to R₁ = 4.51% for 1413 unique reflections. The structure is orthorhombic, Pnma, with lattice parameters a 9.982(1), b 8.502(1), c 4.5480(6) Å, Z = 4, and isotypic with norbergite, Mg₃SiO₄F₂. The empirical-chemical formula, based on the electron-microprobe determination of elements heavier than oxygen, and beryllium tentatively calculated to give 1Be for O = 6, is Be(V³⁺_1.27Ti_1.21Cr_0.29Fe_0.08Al_0.07)_S2.92O₆. The presence of Be was verified by the structure analysis and a SIMS analysis. The occupancy of the Be site was found to be 0.84(1), which largely explains the surplus of Ti in the empirical formula, as compared to the stoichiometric formula BeV₂TiO₆. However, there is no evidence for ordering of V and Ti on two unique octahedral sites of the structure. The amended empirical formula is (Be_0.84¿_0.16)(V³⁺_1.32Ti_1.25Cr_0.29Fe_0.09Al_0.07)_S3.02O₆.  Appearance of the mineral, the dominance of V and Ti in its composition and the X-ray powder-diffraction data correspond to those described originally for kyzylkumite, supposed to be V₂Ti₃O₉. The crystal-lattice parameters found earlier for kyzylkumite can be explained  as based on an 8x supercell obtained on a {210} reflection twin. This type of twinning is characteristic for the Byrud mineral, and was present in the crystal used for the structure determination. These facts suggest that the same type of oxide of Be, V and Ti found at Byrud was also the main constituent of material described as kyzylkumite, which probably also contained additional phases, and that a redefinition of this mineral is needed.