A laboratory-assembled laser-induced breakdown spectroscopy instrument consisting of a compact low-power diode-pumped solid-state laser and a palm-sized low-resolution non-gated spectrometer was evaluated for the quantification of lithium in lithium-bearing ore. Five ore materials containing 0.468 – 2.67 wt.% lithium were analyzed. Three Li I emission peaks at 610, 670, and 812 nm were used after total-intensity normalization. The total-intensity normalization produced a modest improvement in intra-pellet precision but a much larger improvement in calibration accuracy by effectively reducing inter-pellet variation. All three peaks yielded good univariate linear calibration. Among the unconstrained single-peak models, the 670 nm peak showed the lowest prediction error, whereas the 812 nm peak became the best univariate variable when a zero-intercept model was applied, owing to its near-ideal sensitivity and minimal self-absorption. Multilinear regression showed that the combination of the 610 and 812 nm peaks gave the best prediction performance in leave-one-ore-out cross-validation, outperforming all univariate models. In contrast, the use of all three variables did not improve the model further because of strong multicollinearity among the Li peak intensities. These results demonstrate that reliable lithium quantification can be achieved with a low-cost LIBS instrument and that careful selection of a small number of physically meaningful and complementary spectral variables is more important than simply increasing model complexity.