Mica is the most important carrier of rare alkalis (Li, Rb and Cs) and volatiles, and also a major host of rare metals such as Sn, W, Nb and Ta. Thus, the knowledge of trace element partitioning between mica and silicate melt is crucial to understand enrichment mechanisms of ore metals in evolved silicic systems. However, experimental data are scarce, mainly because the synthesis of large mica grains which can be analyzed with laser ablation techniques is challenging. To address this issue, we conducted two-step experiments, using a high-temperature step at 800 °C, and a low-temperature step at 600 °C with a temperature cycling interval of 30 °C, and variable run durations from 21 to 60 days. The experimental products, at 600 °C, 200 MPa, are mainly composed of mica and alkali feldspars. Ferroholmquistite (Fhlm; a Li-bearing amphibole) was also observed in the Li-bearing systems.Beryllium, Zn and Cs are highly incompatible in all run products with mineral/melt partitioning of mineral/meltDBe between 0.01 and 0.08, mineral/meltDZn between < 0.01 and 0.24 and mineral/meltDCs between <0.01 and 0.19. Gallium is mildly incompatible in mica and feldspars with mica/meltDGa, K-feld/meltDGa and Na-feld/meltDGa of 0.24–0.35, 0.60–0.67 and 0.69–0.72, respectively, whereas Fhlm/meltDGa is nearly 1. Rubidium is characterized by a slight enrichment in mica and K-rich feldspar with respect to melt. Niobium and Ta are highly incompatible in feldspars, while they are compatible in ferroholmquistite. Li-mica crystallization would cause Nb/Ta fractionation with mica/meltDNb and mica/meltDTa of 4.46 and 0.88, respectively. Tungsten is preferentially incorporated in Li-rich mica with mica/meltDW > 22. The strong partitioning of W into Li-rich mica implies that minor Li-rich mica fractionation (>5 %) will strongly deplete W in the residual melt, thus inhibiting transfer to the magmatic fluid at the magmatic-hydrothermal transition. Based on the observed mineral/meltD value, a multi-stage quantitative fractionation model for the behavior of metals during the crystallization of Li-bearing granitic magmas was developed. The results suggest that crystal fractionation is an effective mechanism that could further enrich Be, Ga, Rb, Cs, Nb and Ta in residual melts. Li-rich mica and muscovite are more effective in causing Nb/Ta fractionation as compared with biotite. The very low Nb/Ta ratios found in natural granite may have resulted via extreme, i.e., >99 wt% of fractional crystallization, even if without the precipitation of columbite-group minerals or increasing hydrothermal activity. Due to the very high Li-mica/meltDW value, water saturation in the melt and subsequent magmatic fluid exsolution should occur before Li-mica starts to crystallize to avoid W sequestration in the magma. At 175 MPa, the efficiency of extracting W from Li-poor granitic system is at least 15 % higher than that from Li-rich granitic system. Li-poor granitic systems are thus expected to have a greater W-mineralization potential than Li-rich system, which is consistent with the fact that most parental magmas of large/giant W deposits are Li-poor in composition.