The objective of this study was to verify and compare the P availability in an Oxisol cultivated with Megathyrsus maximus cv BRS-Zuri and Urochloa decumbens cv Basilisk and to predict the best soil P extractor in pastures in Southern Amazonia. Two experiments were carried out, one with BRS-Zuri grass and the other with Signal grass. The experimental design was randomized blocks with five doses of P2O5 based on the maximum capacity of adsorption of P (MCAP), corresponding to 0, 3, 6, 12 and 24% MCAP, with doses 0, 54, 108, 216 and 432 kg ha-1 of P2O5, with four replicates. In each cut of the grasses were determined the dry mass, the accumulation of leaf P and by the Mehlich-1 and resin extractors verified the available P. There were effects of P2O5 doses on the variables analyzed in both grasses. The appropriate class of P by Mehlich-1 was 14.6 mg dm-3 (BRS-Zuri) and 7.2 (U. decumbens) mg dm-3, by the resin extractor was 20.5 mg dm-3 (BRS-Zuri) and 14.40 mg dm-3 (U. decumbens). The resin extractor yielded the highest correlation indexes between PR and soil P content.