Abstract: Glass fiber reinforced polymer (GFRP) anchor is a new type of composite anchor developed from non-metallic anchors. It has the advantages of light weight, high tensile strength, low cost, good corrosion resistance and strong electromagnetic interference resistance. Based on the destructive pull test of GFRP anchors and reinforced anti-floating anchor conducted on a medium-weathered granite site, the displacement of the anchor body and anchor solid during the drawing process of the anti-floating anchor was measured. The bearing capacity of anti-floating anchors with different materials and different anchoring lengths and the relative slip between anchor body and anchor were analyzed. The different load-displacement models were compared and the most suitable load-displacement model for rock anti-floating anchors was sought. The test results show that: In medium-weathered granite, the GFRP anti-floating anchors at the same anchoring length increase the failure load by 13% to 14% compared with the reinforced anti-floating anchors. GFRP anti-floating anchors are more prone to pull-out and failure of the anchor body, and the residual bearing capacity of the anchoring system is still not exerted. It is feasible to use GFRP anchors instead of steel anchors. Compared with the GFRP anti-floating anchor with an anchoring length of 4.5 m, the anchor body with an anchoring length of 6.5 m has a greater slippage relative to the anchor solid. Increasing the anchoring length of the GFRP anti-floating anchor can effectively increase its relative slip, and increasing the anchoring length has no obvious effect on the failure mode of the reinforced anti-floating anchor. The hyperbolic function and power function load-displacement curve models are in poor agreement with the measured values, while the finger-power function curve model has the highest accuracy in predicting the failure load of the anchors in this test, and the overall trend of the curve is more consistent.