Abstract Grass-dominated ecosystems cover ~40% of Earth's terrestrial surface, with tropical grasses accounting for ~20% of global net primary productivity. C 3 (cool/temperate) and C 4 (tropical and subtropical) grass distribution is driven primarily by temperature. In this work, we used phytolith assemblages collected from vegetation plots along an elevation and temperature gradient in the northern Andes (Colombia and Ecuador) to develop a paleothermometer for the region. To accomplish this, we created a transfer function based on the inverse relationship between mean annual temperature (MAT) and the phytolith-based climatic index ( Ic ), which is the proportion of C 3 over C 4 grass phytoliths (GSSCP). To evaluate how accurately the index reflects C 4 –C 3 grass abundance in vegetation plots, we compared it with semiquantitative floristic estimates of C 4 –C 3 grass abundance. To further evaluate the 1 − Ic index as a proxy for C 4 –C 3 grass abundance, we compared it with corresponding δ 13 C values (an independent proxy for C 4 –C 3 vegetation). Results indicate that (1) GSSCP assemblages correctly estimate C 4 –C 3 grass abundance in vegetation plots; (2) the Ic index outperforms the δ 13 C record in estimating C 4 –C 3 grass abundance, even in open-vegetation types; and (3) our Ic index–based model accurately predicts MAT. This new calibrated proxy will help improve paleotemperature reconstructions in the northern Andes since at least the emergence and spread of C 4 grasses in the region during the late Miocene.